Modelling parking behaviour of commercial vehicles: a scoping review

ABSTRACT

Parking in dense urban areas is a major challenge for last mile logistics. Parking shortage and policies that do not address commercial vehicles’ needs often lead these vehicles to park illegally. This paper conducts a scoping literature review on the parking behaviours of commercial freight and service vehicles, methods used to model these behaviours, and factors that determine their outcomes. Thirty-four studies are included in the review. It is found that commercial vehicles’ parking behaviours mainly comprise parking location and type choices including illegal parking, parking duration, and parking cruising. Methods used to model these behaviours primarily include discrete-choice modelling, regression analysis, survival analysis and simulation. We identify key knowledge gaps and provide insights on research opportunities in modelling more complex parking decisions, investigating parking cruising of commercial vehicles, evaluating the implications of freight demand management, and developing data fusion techniques.

KEYWORDS:

• Freight parking
• parking behaviour
• parking choice
• scoping review
• illegal parking
• parking duration

1. Introduction

In dense urban areas, commercial freight vehicles struggle to find appropriate parking (Nourinejad et al., 2014). The resulting delivery delays affect service quality and reliability. Parking is a major challenge for freight operations in urban areas and competition for curb side space frequently results in parking shortage (Girón-Valderrama et al., 2019). Many freight vehicles either park illegally or cruise for parking (Jaller et al., 2013). Commercial freight vehicles often park in curb side space where parking is not allowed or where loading operations are prohibited, for example travel lanes, no stopping zones, bus stops, or cycle lanes.

Parking shortage results when parking infrastructure and policies fail to accommodate the operational requirements of freight vehicles. While passenger vehicles usually park once at trip ends, commercial freight vehicles conduct tours and park at each pickup/delivery stop (Nourinejad et al., 2014). Commercial vehicles often park for short periods, close to their destinations, and require large parking spaces to accommodate their size and loading/unloading activities (Amer & Chow, 2017). While parking restrictions can influence drivers of passenger vehicles to switch to other modes, deliveries by commercial vehicles generally do not have other mode options, leading to illegal parking (Jaller et al., 2013).

Illegal parking of commercial vehicles is pervasive, for example in Toronto, courier companies paid about CAD $27 million in parking fines in 2012 (Haider, 2009; Wenneman et al., 2015). Illegal parking by commercial vehicles also adversely impacts traffic congestion and safety. Han et al. (2005) show that illegal parking by delivery vehicles is among the three leading causes of nonrecurrent congestion in urban areas, behind crashes and construction, estimating that illegally parked delivery vehicles result in an annual delay of around 476 million vehicle-hours and around USD $10 billion in travel time cost in the United States. Illegally parked vehicles jeopardise safety of other sidewalk, curb side, and road users. For example, Conway et al. (2013) show that on average about 14% of curb side commercial vehicle loading events conflict with cyclists.

Understanding parking patterns is needed for planning freight travel activity, the design and management of parking infrastructure, and the development of effective freight demand management and parking policies. Policy measures researched in the literature and adopted in pilots have shown potential in tackling the parking problem of commercial vehicles (Sharman et al., 2012). However, studies are mostly context or location specific and based on qualitative evidence gathered from interviews and focus groups rather than quantified relationships (Wenneman et al., 2015). Data-driven analytical tools would allow for better characterisation of the parking patterns of commercial vehicles.

The literature lacks a comprehensive review of data-driven approaches to modelling the parking behaviours of commercial vehicles. This scoping literature review aims to characterise commercial vehicles’ parking behaviours, explore methods used to analyse these behaviours, understand the factors that influence them, identify knowledge gaps, and guide future research in this field.

The paper is organised as follows: Section 1 is an introduction, Section 2 provides background literature, Section 3 presents the methodology, Section 4 provides parking-related terminology for commercial vehicles, Section 5 presents the review results, Section 6 provides a discussion of the key literature gaps and areas of future research and summarises the review’s limitations.

2. Background literature

Considerable research has been undertaken on modelling of passenger vehicle parking, which is relevant background for this review of commercial vehicle parking. Early research by Austin (1973) models parking of commuters in central business districts with focus on walking distance from parking location to destination and travel cost. Whitlock (1973) proposes an allocation model which distributes parking spaces such that cost of travel is minimised. Similarly, Arnott et al. (1991) develop a network congestion model to investigate the impacts of parking pricing on the spatial and temporal distribution of parking demand.

While these studies assess impacts of parking supply and cost on parking choice, they do not explicitly model parking behaviour. Gillen (1978) addresses this by developing a nested logit model that predicts parking location and household relocation choices in response to parking pricing policies. Van Der Goot (1982) and Hunt and Teply (1993) extend the model developed by Gillen (1978) to illustrate the interactions among the choices of travel mode, parking infrastructure, and parking location. Habib et al. (2012) also incorporate choices of departure time and parking duration. Asakura and Kashiwadani (1994) find that the availability of parking information significantly influences parking choices. Guo et al. (2013) account for parking information uncertainty on drivers’ parking choices using a neo-additive capacity model which considers the variability of drivers’ perception of parking information accuracy. Ottomanelli et al. (2011) also consider uncertainty of the information on parking availability and incorporate factors pertaining to congestion, delay, and enforcement policies.

Simulation models provide additional means to evaluate parking systems and implications of policies under uncertainty. PARKAGENT, a dynamic agent-based simulation model, is used to model the impacts of passenger car parking on traffic congestion (Benenson et al., 2008). SUSTAPARK evaluates how information on parking availability, cost, and accessibility affects the parking location and type choices of commuters (Dieussaert et al., 2009). Waraich and Axhausen (2012) use MATSim to evaluate how parking information impacts commuters travel planning. Ni and Sun (2017) also use agent-based simulation to evaluate the effect of parking reservation systems on parking choice. This study confirms that the provision of timely parking information reduces travel delay and achieves more optimal allocation of parking spaces.

Although there has been considerable progress in the development of quantitative analyses of parking behaviour, almost none of these tools account for commercial vehicles. Commercial vehicles’ travel and parking behaviours significantly differ from those of private passenger vehicles, yet they often compete for the same parking spaces. A complete assessment of parking must account for parking patterns of passenger, and commercial freight and service vehicles. This scoping review describes the existing literature on modelling the parking behaviour of commercial vehicles and provides a summary of the key knowledge gaps warranting further research.

3. Methodology

This review follows the Joanna Briggs Institute (JBI) manual for scoping reviews and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for scoping reviews, PRISMA-ScR (Moher et al., 2015).

3.1. Study search and selection

The literature search was carried out in: Scopus ®, Compendex ®, Web of Science ^TM, ScienceDirect, and Transportation Research International Documentation (TRID) which includes prominent publications specialised in transportation research. A snowball search through the bibliographies of the identified studies was also undertaken.

A search query was developed using three keyword categories: population, outcomes, and methods (see Table 1). The population comprises motorised commercial freight and service vehicles, which use road networks to transport goods and provide services. Outcomes refer to the patterns depicting the parking activity of commercial vehicles. These include choices of parking location and infrastructure type, illegal parking (unauthorised use of infrastructure), parking duration, stop frequency, parking cruising and queuing at parking facilities. Methods include evidence-based analyses which characterise and/or quantify the parking patterns of commercial vehicles. Specifically, we consider studies that deploy data-driven analytical or quantitative empirical research methods to describe, model, or predict parking behaviour. These methods include a wide range of statistical, simulation, and machine learning models.

Table 1. Study search keywords.

Item	Keywords
Method	Model/models/modeling/modelling
Outcome	Parking
	Parking behaviour/behavior
	Parking choice
	Parking location
	Parking type
	Parking dwell time/duration/period
	Loading/unloading zone/area/bay
Population	Commercial vehicle
	Truck
	Van/minivan
	Lorry
	Heavy vehicle/heavy goods vehicle/HGV
	Light goods vehicle/LGV
	Freight vehicle
	Heavy duty vehicle
	Light duty vehicle
	Utility vehicle
	Service vehicle
	Long haul vehicle
	Trailer/tractor/tractor trailer/tractor-trailer

Identified studies were also filtered according to the study format, date, and language. Eligible studies included peer-reviewed and published articles, dissertations and theses, conference proceedings, as well as technical reports from the grey literature. The search only included studies reported in English and published between January 1990 and April 2023, inclusively.

3.2. Data charting and extraction

After removing duplicate records, we conducted title and abstract screening, full-text screening, and data extraction. The results of the study search and selection are illustrated in a PRISMA flow diagram (Figure 1). Thirty-four studies are included in this review.

Figure 1. PRISMA flow diagram of study search and selection processes.

Two independent reviewers carried out title and abstract screening. Preliminary screening questions, search highlights and filters were used to facilitate the process. Next, a more exacting evaluation of the candidate studies was undertaken in the full-text screening stage. Discrepancies between the reviewers on the inclusion/exclusion of studies were resolved by discussion and consensus or by an independent third party when they persisted. Finally, data were manually extracted from the eligible studies, and the information was synthesised to comply with the review’s objectives, identify gaps in the literature, and discuss opportunities for future research.

4. Terms and definitions

We identify four parking behaviours of commercial vehicles: parking location and/or type choice (n = 23), parking duration (n = 14), parking cruising (n = 5), and queuing at parking facilities (n = 2).

We identify two approaches to classifying parking location and/or type: (1) on-street versus off-street, and (2) legal versus illegal. Eighteen studies consider on-street parking only, four consider off-street parking only, and nine consider both. Thirteen studies focus on legal parking, four focus on illegal parking, and 14 consider both. Two studies, Sharman et al. (2012) and Mjøsund and Hovi (2022), do not specify parking type; instead, they define parking as any stopping event whose duration exceeds a presumed cutoff value.

On-street parking consists of curb side spaces that can accommodate commercial vehicles, including commercial vehicle or freight loading zones (CVLZ or FLZ). Alleyways and driveways also provide permissible on-street parking for commercial vehicles. In addition to surface parking lots and parking garages, off-street parking includes loading and unloading bays and highway rest areas which host various services and amenities.

A legal parking space can be characterised as designated or permissible. For example, CVLZs are designated parking spaces, whereas general curb side spaces comprise permissible parking for commercial vehicles. Illegal parking spaces either do not accommodate parking or have usage restrictions. Common illegal parking by commercial vehicles includes double parking and stopping in no parking/stopping/standing zones, transit lanes or stops, fire hydrant lanes, street medians, and highway ramps and shoulders (Anderson et al., 2018; Kawamura et al., 2014; Wenneman et al., 2015). Illegal parking also includes violations of time-of-day, vehicle type, vehicle size, or plate restrictions, as well as expired meter, unpaid fees, or permit violations (Kim & Wang, 2021).

Kawamura et al. (2014), Wenneman et al. (2015), and Kim and Wang (2021) model illegal parking as the number of parking violations by commercial vehicles, and Nourinejad and Roorda (2017) and Abhishek et al. (2021) model the fraction of vehicles engaging in illegal parking. Eleven studies consider illegal parking as a choice alternative or as a determinant of parking behaviours, such as parking duration (Low et al., 2020; Schmid et al., 2018) and parking search time (Dalla Chiara & Goodchild, 2020; Holguin-Veras, 2016).

Of the reviewed studies, 14 studies investigate the determinants of the parking duration of commercial vehicles. Parking duration is the time a vehicle remains stationary for the purpose of conducting an activity or resting. For example, Dalla Chiara and Cheah (2017) model the parking facilities of an urban retail mall as a queuing system. Three time periods are distinguished: dwell time, queuing time, and parking duration, whereby dwell time is the summation of the queuing time and parking duration. Parking duration of commercial freight vehicles is related to the stopping frequency, which is the number of pickup and/or delivery stops in a vehicle tour (Sharman et al., 2012; Zou et al., 2016).

Five studies analyse the parking cruising behaviour of commercial vehicles, which is characterised by the search time, i.e. the time until a vehicle finds the first designated or permissible, vacant parking space. Dalla Chiara and Goodchild (2020) and Dalla Chiara et al. (2022) quantify parking search time as a trip length deviation, i.e. the difference between the actual trip length recorded by GPS (Global Positioning System) device and expected trip length as estimated by Google Maps ®. Holguin-Veras (2016), Lopez et al. (2019), and Dalla Chiara et al. (2022) use parking search time as a metric to evaluate policies on CVLZs.

While parking cruising can be thought of as “invisible queues”, queuing at parking facilities involves visibly awaiting a parking spot and is directly related to parking occupancy. Dalla Chiara and Cheah (2017) and Dalla Chiara et al. (2017) model the parking facilities of a large retail mall as a queue system. The queue time is quantified as the time spent waiting for the first loading/unloading bay to become vacant. Freight vehicles comprise arrivals to the queue system, loading/unloading bays are servers, and service time is equal to the parking duration. Dalla Chiara and Cheah (2017) model queue time as a function of queue length and consider it a determinant of parking location/type choice. In Dalla Chiara et al. (2017), the queue time is a simulation output metric to evaluate the effectiveness of implementing centralised goods receiving policies at the mall’s loading/unloading bays.

5. Results

5.1. Research trends and origins

All included studies, except Munuzuri et al. (2002), were published in the last decade. Eighteen studies originate from North America, of which 12 are from the United States and six are from Canada. Eight studies originated from Western Europe, seven from East Asia, six of which are from Singapore, and one study from Australia (see Table 2).

Table 2. General characteristics of the included studies.

Study	Country	Study setting	Outcome(s)	Method(s)
Munuzuri et al. (2002)	Spain	Hypothetical Manhattan-style network	Parking location/type	Simulation
Sharman et al. (2012)	Canada	Greater Toronto & Hamilton Area	Parking duration	Survival analysis
Kawamura et al. (2014)	United States	City of Chicago	Illegal parking	Regression analysis
Nourinejad et al. (2014)	Canada	Downtown Toronto	Parking location/type	Discrete-choice modelling Simulation
Gatta and Marcucci (2014)	Italy	Limited traffic zone in city of Rome	Parking location/type	Discrete-choice modelling
Wenneman et al. (2015)	Canada	Downtown Toronto	Illegal parking	Regression analysis
Marcucci et al. (2015)	Italy	Limited traffic zone in city of Rome	Parking location/type	Discrete-choice modelling
Rosenfield et al. (2016)	Canada	City of Toronto	Parking location/type	Discrete-choice modelling
Gatta and Marcucci (2016)	Italy	Limited traffic zone in city of Rome	Parking location/type	Discrete-choice modelling
Zou et al. (2016)	United States	Manhattan Area	Parking duration Stopping frequency	Survival analysis Regression analysis
Holguin-Veras (2016)	United States	Midtown Manhattan	Parking search time Illegal parking	Simulation
Dalla Chiara et al. (2017)	Singapore	Urban retail mall	Queue time at loading bays Illegal parking	Simulation
Amer and Chow (2017)	Canada	Downtown Toronto	Illegal parking	Equilibrium analysis
Nourinejad and Roorda (2017)	Canada	Downtown Toronto	Illegal parking	Inspection game modelling
Dalla Chiara and Cheah (2017)	Singapore	Urban retail mall	Parking duration Queue time at loading bays	Regression analysis
Anderson et al. (2018)	United States	Highways in Pacific Northwest region	Illegal parking	Discrete-choice modelling
Schmid et al. (2018)	United States	City of New York	Parking duration	Survival analysis
Girón-Valderrama et al. (2019)	United States	Downtown Seattle	Parking location/type Parking duration	Quantitative analysis
Lopez et al. (2019)	France	City of Lyon	Parking search time	Simulation
Low et al. (2020)	Singapore	Urban retail mall	Parking duration	Machine learning
Mahmud et al. (2020)	United States	State of Arkansas inter-city highways	Parking duration	Machine learning
Gopalakrishnan et al. (2020)	Singapore	Singapore	Parking location/type	Discrete-choice modelling Simulation
Seya et al. (2020)	Japan	Sanyo & Chugoku Expressways	Parking duration	Regression analysis
Dalla Chiara et al. (2020)	Singapore	Urban retail mall	Parking location/type	Discrete-choice modelling Simulation
Dalla Chiara and Goodchild (2020)	United States	Downtown Seattle	Parking search time	Regression analysis
Dalla Chiara et al. (2021)	United States	City of Seattle	Parking location/type Parking duration Parking search time	Quantitative analysis
Kim et al. (2021)	United States	Urban mixed-use developments	Parking duration	Regression analysis
Kim and Wang (2021)	United States	City of New York	Illegal parking	Regression analysis
Abhishek et al. (2021)	Australia	Hypothetical network of urban streets	Illegal parking	Queue analysis
Castrellon et al. (2023)	Spain,	Freight loading zones in the city of Vic	Parking duration	Machine learning
Alho et al. (2022)	Singapore	Urban retail mall	Parking location/type Parking duration	Simulation
Dalla Chiara et al. (2022)	United States	City of Seattle	Parking search time	Regression analysis
Kalahasthi et al. (2022)	Spain	Freight loading zones in the city of Vic	Vehicle arrival rate Parking duration	Regression analysis Survival analysis
Mjøsund and Hovi (2022)	Norway	Cities of Oslo, Bergen, Trondheim, Stavanger, Kristiansand, Drammen & Tromsø	Parking duration	Quantitative analysis

5.2. Modelling commercial vehicles’ parking behaviours

The methods used by the reviewed studies to describe, model or predict the parking behaviours of commercial vehicles include regression analysis (n = 10), discrete-choice modelling (n = 8), simulation (n = 8), survival analysis (n = 4), and other methods like quantitative empirical analysis (n = 3), machine learning (n = 3), queue analysis (n = 1), equilibrium modelling (n = 1), bilateral search and match (n = 1), and controlled experimentation (n = 1).

5.2.1. Regression analysis

Three studies use log/linear regression: Dalla Chiara and Cheah (2017) and Alho et al. (2022) model the parking duration of commercial freight and service vehicles at a large retail mall located in Singapore, and Dalla Chiara and Goodchild (2020) model the parking search time of parcel delivery vehicles in downtown Seattle. To account for random variations across drivers and locations, Dalla Chiara and Goodchild (2020) propose a mixed-effect loglinear regression model. Dalla Chiara et al. (2022) use a similar model to estimate parking search time of freight vehicles in Seattle before and after providing real-time information on CVLZ parking availability.

Kim et al. (2021) deploy generalised linear regression to explain the parking duration of commercial vehicles delivering goods to large, urban traffic generators in Seattle. Kalahasthi et al. (2022) model arrivals of freight vehicles at FLZs in Vic, Spain as a negative binomial process. Also, Zou et al. (2016) use generalised linear regression to model the stopping frequency of delivery trucks in Manhattan, New York City.

Kawamura et al. (2014) and Kim and Wang (2021) use generalised linear regression to model illegal parking frequency of commercial vehicles. Specifically, Kim and Wang (2021) estimate a one-inflated Poisson which reflects that a commercial vehicle engages in illegal parking more frequently than a passenger vehicle as observed. Wenneman et al. (2015) propose a distance-decay weighted regression model to explore the spatial dependency of illegal parking frequency on parking demand and supply, and the built environment.

Seya et al. (2020) use a Type II Tobit model to predict stopping and stop durations at highway rest areas. This is a form of censored regression analysis whereby a selection model predicts whether a truck stops at a rest area, and the output model predicts the parking duration.

5.2.2. Discrete-choice modelling

Discrete-choice models are used to explain the decision-making process, particularly the dependence of decisions on the attributes of the decision maker and available alternatives (see Table 3). The most prevalent model specification is the multinomial logit model (MNL), of which a special case is the binary logit model (Nourinejad et al., 2014).

Table 3. Basic components of discrete-choice models developed by included studies.

Study	Decision(s)	Decision-maker(s)	Choice set	Model(s)
Nourinejad et al. (2014)	Parking location	Vehicle driver	Park in available location Do not park in available location	BNL
Gatta and Marcucci (2014)	Loading bay policy	Owner-operator Retailer Transport provider	Loading/unloading bay policy (capacity, availability & fee)	MNL
Marcucci et al. (2015)	Loading bay policy	Transport provider	Loading/unloading bay policy (capacity, availability & fee)	MNL MN-LC
Rosenfield et al. (2016)	Purchasing a parking permit	Vehicle owner	Purchase parking permit Do not purchase parking permit	NL
	Parking type	Vehicle driver	Legal parking Illegal parking
Gatta and Marcucci (2016)	Loading bay policy	Retailer Transport provider	Loading/unloading bay policy (capacity, availability & fee)	MNL
Anderson et al. (2018)	Issue with finding parking	Vehicle driver	Issues finding parking No issues finding parking	ML
Gopalakrishnan et al. (2020)	Overnight parking location & type	Vehicle owner	Public parking lot Open-access private parking lot Closed-access private parking lot	MNL
Dalla Chiara et al. (2020)	Parking type	Vehicle driver	Loading/unloading bay Mall's car park Double parking	MNL ML
Alho et al. (2022)	Parking type	Vehicle driver	Loading/unloading bay Mall's car park Double parking	MNL

Rosenfield et al. (2016) propose a nested logit model to evaluate the financial feasibility of a parking permit program which would allow commercial vehicles to park in no parking, standing, or stopping zones in Toronto. A hierarchal decision-making structure is considered; at the upper level, vehicle fleet owners decide whether to purchase a permit, and at the lower level, drivers without permits choose legal or illegal parking.

To account for heterogeneity among decision-makers, it is common for logit models to be segmented by one or more characteristics of the decision-makers. Dalla Chiara et al. (2020) segment vehicle drivers by vehicle size, observe behavioural differences and incorporate random taste variation using a mixed logit (ML) model specification. The ML resolves two other limitations of the MNL: restricted substitutions across choice alternatives and time-dependent correlations between unobserved factors. Random taste variation among truck drivers is also considered in Anderson et al. (2018).

Gatta and Marcucci (2014) and (2016) segment decision-makers by agent type, into owner-operator, retailer, and transport provider. A multinomial logit latent class model (MN-LC) is estimated in Marcucci et al. (2015) to account for intra-agent heterogeneity. Because the objective is to evaluate policy effectiveness, the authors segment transport providers into classes based on observed behavioural differences.

5.2.3. Simulation

The simulation models developed by the reviewed studies fall into two categories, discrete-event (n = 4) and agent-based (n = 4) simulation. While discrete-event simulation models the progression of temporal events, agent-based simulation focuses on the interactions between agents such as vehicle drivers, establishments, and carriers. Table 4 summarises characteristics of the simulation models, including parking resources, scenarios, and the metrics used to assess scenarios.

Table 4. Characteristics of simulation models developed by included studies.

Study	Model	Parking resources	Scenarios	Metrics
Munuzuri et al. (2002)	Agent-based: Hypothetical city-wide traffic model	General on-street parking spaces, cruising, loading bays & double parking	None	Parking assignments
Nourinejad et al. (2014)	Discrete event: Pear hour downtown traffic model	General on-street parking & loading bays	Restricted/unrestricted freight access and parking	Parking assignments Average parking search time Average access time Average walking time Total network travel time
Holguin-Veras (2016)	Discrete event: Regional freight traffic model	General on-street parking, loading bays & double parking	Increase/decrease parking supply Increase/decrease parking duration	Average parking search time Average walking distance
Dalla Chiara et al. (2017)	Discrete event: Queue system of retail mall's parking facilities	Loading bays, off-street garage & double parking	Efficient/inefficient goods handling	Total queue time at loading bays Fraction of vehicles parked outside loading bays
Lopez et al. (2019)	Agent-based: City-wide traffic model	Freight loading zones	None	Average parking search time
Gopalakrishnan et al. (2020)	Agent-based: City-wide freight traffic model	Overnight public or private parking facilities	Enhanced distribution of overnight parking supply	Parking assignments Vehicle tours Total vehicle-kilometres travelled
Dalla Chiara et al. (2020)	Discrete event: Queue system of retail mall's parking facilities	Loading bays, off-street garage & double parking	Less/more loading bays Free/paid parking Increase/decrease parking fine Low/high enforcement level Efficient/inefficient goods handling	Parking assignments Average queue time at loading bays Average delivery cost Total CO2 emissions
Alho et al. (2022)	Agent-based: City-wide traffic model	Loading bays, off-street garage & double parking	Increase capacity of loading bays Free parking Prohibit unauthorised parking Efficient goods handling Enhanced parking information & guidance	Parking assignments Parking duration Total queue length & time at parking facilities Total cost of travel & waiting time Total revenue from parking fees Total HGH emissions

5.2.4. Survival analysis

Survival analysis involves estimating a hazard model of duration as a function of explanatory variables. The hazard function predicts the limiting probability of the departure time from a parking location. Hazard models can be parametric, semi-parametric, and non-parametric. Among the reviewed studies, four studies use survival analysis to predict the parking duration of commercial vehicles.

Schmid et al. (2018) use a Weibull parametric hazard model to model the parking duration of commercial vehicles delivering goods and services in Manhattan New York. A similar model specification is adopted by Kalahasthi et al. (2022) to depict the parking duration of vehicles at freight loading zones in Vic, Spain. Zou et al. (2016) opt for a semi-parametric hazard model with a Cox-proportional specification, to estimate the parking duration of delivery trucks.

Sharman et al. (2012) compare the performance of parametric and non-parametric hazard models in estimating stop duration of commercial vehicles. The study tests an accelerated failure time parametric hazard model with log logistic distribution and a proportional non-parametric hazard model with discretised stop duration intervals. Although signs and magnitudes of estimated parameters follow similar patterns, the parametric model provides a better fit than its non-parametric counterpart.

5.2.5. Other methods

Girón-Valderrama et al. (2019) and Dalla Chiara et al. (2021) use on-site shadowing of parking events and ride-along surveys, respectively, to characterise the parking patterns of commercial vehicles, including location and infrastructure type, duration, and search time. While Girón-Valderrama et al. (2019) observe freight parking activity around large freight traffic generators in downtown Seattle, Dalla Chiara et al. (2021) record parking-related decisions of drivers conducting deliveries across Seattle. Mjøsund and Hovi (2022) use GPS records of the tours travelled by freight vehicles across seven Norwegian cities to map the parking activity. They develop an approach to identify and classify delivery stops and estimate parking durations.

Low et al. (2020), Mahmud et al. (2020), and Castrellon et al. (2023) use machine learning algorithms to predict parking durations and usage patterns of commercial vehicles. Low et al. (2020) and Castrellon et al. (2023) use gradient boosting algorithms to predict parking durations of commercial delivery vehicles at a large urban retail mall and FLZs. Mahmud et al. (2020) use a bi-level unsupervised learning algorithm to predict truck parking usage of highway rest areas by service type and duration.

Abhishek et al. (2021) model curb side parking as a queue system in which a street segment has a parking lane comprising general on-street parking spaces and freight loading bays. While passenger cars can only use general on-street parking spaces, commercial vehicles can use both parking provisions; therefore, arrivals to on-street parking spaces include passenger vehicles and commercial freight vehicles that have found all bays occupied. Accordingly, illegal parking by commercial freight vehicles is assumed for vehicles denied usage of any parking provision.

Amer and Chow (2017) extend the parking and traffic congestion model developed by Arnott and Inci (2006) by incorporating the parking behaviour of delivery trucks. Delivery trucks are assumed to only park in designated on-street parking spaces, otherwise they would resort to double parking. Equilibrium analysis is undertaken to study the interaction between passenger cars and delivery trucks and to explore how the parking behaviour of the latter affects travel time, speed, and the parking cruising of passenger cars.

Lastly, Nourinejad and Roorda (2017) use inspection game theory to model the interaction between enforcement units and illegally parked commercial vehicles. The study uses bilateral searching and matching to evaluate how changes in enforcement frequency and fines affect the parking behaviour of commercial vehicles.

5.3. Parking behaviour determinants

Parking behaviour determinants fall into four categories: decision-maker, vehicle, and trip attributes; parking location attributes; activity and shipment attributes; and built environment attributes. Table 5 summarises the effects of key determinants considered by the included studies. Where an effect is quantifiable, upward (↑) and downward (↑) arrows indicate a marginal increase and decrease in parking behaviour outcome, respectively. If a factor is specified as a categorical nominal variable, and found significant with respect to the outcome, it is indicated using a check mark (✓) symbol.

Table 5. Summary of the effects of parking behaviour determinants.

		Parking behaviour outcome
	Determinant	Illegal parking	Legal parking	Parking duration	Parking Cruising	Queue time	Stopping frequency
Driver, vehicle & trip	Vehicle size	↑		↑			↑
	Vehicle age	↑
	Vehicle load		↑	↑			↓
	Trip length			↑	↑
	Time-of-day	↑		✓			✓
	Freight trips	↑
	Day-of-Week			✓
	No. of stops			↓
	Industry sector	✓		✓
	Facility usage frequency		↑	↑
	Distance to destination		↓	↓
	Distance to depot		↓
Parking alternative	Parking type		✓	✓
	Parking cost		↓
	Designated parking capacity/availability	↓	↑		↓	↓
	Permissible parking capacity/availability	↓		✓	↓	↓
	Unauthorised parking curb length	↑			↑
	Facility amenities availability		↑	✓
Activity & shipment	Activity time	↓	↑		↑	↑
	Activity purpose		✓	✓
	Shipment type		✓	✓			✓
	Shipment size	↓	↑	↑
	Deliveries per stop	↓	↑	↑
Built environment	Establishment sales	↑		↑
	Establishment count	↑		↑
	Employment	↑
	Built footprint	↑
	Resident vehicle ownership	↓
	Residential turnover	↑

5.3.1. Decision-maker, vehicle, and trip

In the reviewed studies, common decision-maker, vehicle, and trip attributes are commercial vehicle size, time-of-day of vehicle arrival, and industry sector. Vehicle size is a function of the vehicle’s body type; common commercial vehicles include vans, light or single-unit trucks, heavy trucks, and tractor-trailers. For example, smaller commercial vehicles are associated with shorter parking durations (Dalla Chiara & Cheah, 2017; Kim et al., 2021; Schmid et al., 2018; Zou et al., 2016). Kim and Wang (2021) also find that vans and trucks engage more frequently in illegal parking compared to smaller commercial vehicles. This is in line with Dalla Chiara et al. (2020), who show that heavier commercial vehicles park illegally more often.

Sharman et al. (2012) and Zou et al. (2016) find that time-of-day of a parking event is significantly correlated with the parking duration. Both conclude that stops occurring in the morning tend to be longer. Low et al. (2020) and Mahmud et al. (2020) also conclude that time-of-day is an important input feature to predict parking duration. Sharman et al. (2012) and Seya et al. (2020) show that there is a significant correlation between industry sector and parking duration, while Dalla Chiara et al. (2020) and Gopalakrishnan et al. (2020) show that there is a significant correlation between industry sector and the choices of parking type and location, respectively.

Other decision-maker, vehicle, and trip attributes include vehicle load, trip length, distances to destination and depot, facility usage frequency, day-of-week of occurrence of parking event, vehicle age, and freight travel demand. Dalla Chiara and Cheah (2017) show that higher vehicle load is associated with longer parking duration, and Zou et al. (2016) also show that it is associated with lower stopping frequency. A greater trip length, represented as the travel distance or travel time of trips to or from the parking location, is associated with longer parking duration (Dalla Chiara & Goodchild, 2020; Sharman et al., 2012; Seya et al., 2020).

Abhishek et al. (2021) and Wenneman et al. (2015) evaluate the impacts of freight travel demand on illegal parking by commercial vehicles. Abhishek et al. (2021) show that increasing freight vehicle trips increases the queuing at designated parking facilities, resulting in more illegal parking. On the contrary, Wenneman et al. (2015), find that the number of peak commercial vehicle trips does not have a statistically significant influence on illegal parking frequency.

5.3.2. Parking location

The most prevalent parking alternative attributes are parking type, cost, and capacity. Schmid et al. (2018) discuss the effect of parking type choice on the parking duration of commercial vehicles, showing that commercial vehicles parked illegally tend to have shorter parking durations to avoid citations. Nourinejad et al. (2014) also find that designated parking facilities like loading and unloading bays are preferred over other permissible parking spaces since they better accommodate the size and activity requirements of commercial vehicles. Mahmud et al. (2020) find that overnight parking is more common in rest areas that host enhanced facilities beyond the basic amenities, and Gopalakrishnan et al. (2020) show that heavy commercial vehicles favour controlled access and private lots over open or public access lots for overnight parking.

Parking capacity or availability is a key parking behaviour determinant. More designated and/or permissible parking spaces result in less commercial vehicle cruising (Dalla Chiara & Goodchild, 2020; Holguin-Veras, 2016; Lopez et al., 2019), reduction in average queue time (Dalla Chiara & Cheah, 2017), higher parking utilisation (Dalla Chiara et al., 2020; Gatta & Marcucci, 2014, p. 2016; Marcucci et al., 2015), and reduced illegal parking (Abhishek et al., 2021; Amer & Chow, 2017; Holguin-Veras, 2016; Wenneman et al., 2015). Increased capacity of highway rest areas is also correlated with reduced difficulty in finding safe and adequate parking (Anderson et al., 2018).

Parking cost, in the form of fees and/or fines, generates a disutility for decision-makers. Expensive designated parking facilities result in lower utilisation and more illegal parking (Dalla Chiara et al., 2020; Gopalakrishnan et al., 2020). Gatta and Marcucci (2014, 2016) and Marcucci et al. (2015) show that transport providers and retailers negatively perceive pricing access to the Limited Traffic Zone in Rome, Italy; however, retailers are found to be more price sensitive than transport providers. Lastly, Rosenfield et al. (2016) show that the adoption rate of parking permits for commercial vehicles drops if the permit price exceeds its optimal value.

5.3.3. Activity and shipment

Activity and shipment attributes characterise the transactions undertaken while a commercial vehicle is parked. The main activity and shipment attributes considered in the reviewed studies are activity time and purpose, commodity type, and shipment type and size. Activity time is positively correlated with parking search time and illegal parking frequency by commercial vehicles (Dalla Chiara et al., 2017; Holguin-Veras, 2016). Dalla Chiara et al. (2017) show that an increase in the activity time increases queuing at loading bays in retail malls. Activity purpose is also shown to significantly influence the parking duration of commercial vehicles. Schmid et al. (2018) find that commercial service vehicles are likely to have longer parking durations and Dalla Chiara et al. (2020) show that service vehicles are less sensitive to parking cost. Dalla Chiara and Cheah (2017) and Low et al. (2020) show that when pickup of goods is involved in addition to delivery, the parking duration tends to be longer.

Zou et al. (2016) show that food deliveries have shorter parking durations and higher stopping frequency, whereas Kim et al. (2021) find that commodity type is statistically insignificant to parking duration. Dalla Chiara and Cheah (2017) and Kim et al. (2021) also conclude that larger shipment size has a positive effect on parking duration. Similarly, Dalla Chiara et al. (2020) and Low et al. (2020) show that shipment size influences the parking choice of commercial vehicles at retail malls. Larger shipments often require special handling, and so it is preferrable to use loading/unloading bays. Anderson et al. (2018) also find that shipment size is correlated with truck drivers’ perception of higher difficulty in finding safe and adequate parking on highways.

5.3.4. Built environment

Attributes of the built environment and population include land use patterns and the socio-economic characteristics of the population. Attributes include population density, number of establishments and their types and sales volumes, employment and built footprint. The number, types, and sales volume of establishments are correlated with the parking behaviour of commercial vehicles. Kawamura et al. (2014) and Sharman et al. (2012) show that greater sales volumes are associated with higher illegal parking frequency and longer stop durations, respectively. Kawamura et al. (2014) and Wenneman et al. (2015) also show that a larger number of establishments is associated with more illegal parking. Finally, the size of built footprint is found to be positively correlated with illegal parking frequency by commercial vehicles (Kawamura et al., 2014).

5.4. Policy applications

Parking policies can be grouped into four categories: space management (n = 8), time management (n = 4), pricing (n = 4), and enforcement (n = 3).

5.4.1. Space management

Space management involves the allocation and/or reservation of parking spaces per user type. Amer and Chow (2017) investigate the optimal allocation of parking spaces between passenger vehicles and light delivery trucks. They find that increasing the supply of curb side parking in Toronto’s Central Business District by about 18% and allocating approximately 3.5% of curb side parking to delivery trucks eliminates double parking by delivery trucks and cruising by passenger vehicles. Similarly, Nourinejad et al. (2014) show that increasing the number of designated and permissible curb side parking spaces accessible to freight vehicles results in over 60% reduction in the average search time.

Dalla Chiara et al. (2020) and Alho et al. (2022) show that increasing the number of loading and unloading bays results in the reduction of average delivery cost and total carbon dioxide emissions. The studies illustrate that enhanced capacity improves the utilisation of designated truck parking spaces and reduces illegal parking and idle waiting in queues. Holguin-Veras (2016) evaluates the impacts of enhancing the capacity and availability of loading/unloading bays on parking search time and walking distance for freight vehicle operators. The study shows that increasing the number of loading bays by 5% could result in almost a 61% reduction in average search time of 28 min, and a10% increase in the number of loading bays results in a decrease in the fraction of freight vehicles engaging in illegal parking from 51% to 25%.

Gopalakrishnan et al. (2020) evaluate the impacts of moving overnight parking facilities of heavy commercial vehicles closer to industrial areas. This results in a more balanced city-wide distribution of parking supply despite increasing the lengths of the first and last trips by about 36%, which indicates an increase in the overall operational cost. Nonetheless, this policy can contribute to freeing urban land for more efficient uses and separating heavy commercial vehicles from other traffic, which improves road safety and alleviates congestion.

Dalla Chiara et al. (2022) and Alho et al. (2022) explore the benefits of providing parking information to commercial vehicle drivers. Dalla Chiara et al. (2022) show that providing real-time information on the availability of CVLZs reduces parking cruising by freight vehicles and travel time delay. Alho et al. (2022) show that in addition to providing parking availability information, adopting an effective parking guidance system enhances the utilisation of the parking facilities of large freight traffic generators, and reduces illegal parking by commercial vehicles.

5.4.2. Time management

Time restrictions involve segregating the utilisation of parking spaces by time of day and vehicle type or enforcing time limits on usage. Dalla Chiara and Cheah (2017), Dalla Chiara et al. (2020), and Alho et al. (2022) examine the effects of varying the handling time at loading and unloading bays. More efficient goods handling reduces the parking duration and increases the turnover rate, allowing for improved utilisation of loading bays. Furthermore, adopting a centralised goods receiving policy reduces potential queueing at loading bays, and hence encourages freight vehicles to park legally. Similarly, Holguin-Veras (2016) shows that reducing the handling time at loading bays by 10%, from 20 min on average per vehicle, reduces the average search time by up to 53% from 28 min.

5.4.3. Parking pricing

Dalla Chiara et al. (2020) and Alho et al. (2022) show that free-of-charge loading bays are inefficient due to the increase in induced demand and subsequent queueing, whereas expensive loading bays force more vehicles to park illegally. In Gatta & Marcucci (2014, 2016) and Marcucci et al. (2015), the effects of varying the annual entrance fee to Rome’s Limited Traffic Zone on the utilisation of on-street loading/unloading bays are explored. They show that retailers are more sensitive to cost than transport providers, who usually transfer this component of operational cost to receivers. Therefore, transport providers are more tolerant of reasonable increases in the entrance fee, unlike retailers. Rosenfield et al. (2016) show that the willingness to participate in a parking permit program in downtown Toronto is contingent on how it is priced and the perception of this price by carriers. Higher permit prices cause more carriers to opt for maintaining current parking behaviour patterns and become less averse to risk of citation and fines.

5.4.4. Enforcement

Enforcement level is a function of the inspection frequency and fine amount. Nourinejad and Roorda (2017) investigate the impacts of increased enforcement level in Toronto City. They show that the citation probability is greater for longer dwell times and higher levels of enforcement and that increasing fines and/or level of enforcement reduces illegal parking behaviour. Similarly, Dalla Chiara et al. (2020) and Alho et al. (2022) show that excessive enforcement of curb side parking in the vicinity of retail malls can result in longer queues and delays since they induce a higher demand for designated and permissible parking.

5.5. Data sources

Five primary sources of data are identified in the reviewed studies: on-site observations and inventory survey (n = 14), revealed and stated preference surveys (n = 11), passive data such as GPS and digital tachograph data (n = 6), and citations databases (n = 3).

Studies by Dalla Chiara et al. (2017), Dalla Chiara and Cheah (2017), Low et al. (2020), Dalla Chiara et al. (2020), and Alho et al. (2022) deploy a synthetic dataset characterising the parking activity of commercial vehicles at large urban retail malls in Singapore. This dataset was the result of the fusion of counts from automatic road-side video recordings and the responses of a driver-intercept survey. Low et al. (2020) apply data imputation to this dataset to complete missing fields and predict the parking duration of commercial vehicles. Castrellon et al. (2023) and Kalahasthi et al. (2022) use automatic parking counts to develop parking occupancy models for FLZs in Vic, Spain. Also, Holguin-Veras (2016) use time-lapse cameras to capture the utilisation of curb side space by freight vehicles, inclusive of loading and loading bays.

Nourinejad et al. (2014), Zou et al. (2016), Schmid et al. (2018), Girón-Valderrama et al. (2019), and Kim et al. (2021) conduct on-site shadowing of parking activity of commercial vehicles. While Nourinejad et al. (2014) used observations to develop and estimate a parking location choice model, the other studies focused on explaining the factors influencing the parking duration of commercial vehicles. Observation via ride along surveys were conducted by Dalla Chiara et al. (2021), whereby researchers accompanied drivers on their tours to observe travel and parking behaviours.

Gatta & Marcucci (2014, 2016) and Marcucci et al. (2015) assess the effectiveness of policies on the management of loading/unloading bays within the Limited Traffic Zone of Rome, Italy. They conducted a stated-preference survey by interviewing retailers, transport providers, as well as vehicle owners and operators. Similarly, Holguin-Veras (2016) interviewed truck drivers to collect feedback on parking management policies for loading and unloading bays. Anderson et al. (2018) and Nourinejad et al. (2014) conduct revealed-preference surveys with truck drivers. Anderson et al. (2018) develop a discrete-choice model to predict whether a truck driver would have trouble finding a safe and adequate parking on inter-city highways in Arkansas, United States. Nourinejad et al. (2014) carry out a driver-intercept survey to develop a parking location choice model for delivery trucks in downtown Toronto.

Sharman et al. (2012), Mahmud et al. (2020), and Mjøsund and Hovi (2022) use passive data to model parking duration of freight trucks, while Dalla Chiara and Goodchild (2020) and Dalla Chiara et al. (2022) use similar data to model parking search time of commercial vehicles. These studies use GPS traces of the vehicles travelled paths to identify and characterise the parking patterns of commercial vehicles. Seya et al. (2020) also use passive data retrieved from in-vehicle digital tachographs of trucks to examine the factors influencing the parking durations at highway rest areas.

Kawamura et al. (2014), Wenneman et al. (2015), ad Kim and Wang (2021) use parking citations databases to study illegal parking behaviour by commercial vehicles in Chicago, Toronto, and New York, respectively. The citations databases are fused with databases on the characteristics of the built environment, population, and parking infrastructure.

6. Discussion and conclusion

The results of the scoping review show that research on modelling the parking behaviours of commercial vehicles is limited to only 34 relevant studies. More research is warranted, and, in this section, we identify the main gaps and explore areas for further research.

6.1. Parking choices: the result of complex decision-making

Parking choice is the result of complex decision-making. Discrete-choice models are mostly context-specific, and challenges include transferability and scalability. Study setting particularities and variations among decisions makers give rise to these challenges. Modelling parking choices depends on the study context. For example, the parking patterns of last-mile freight logistics significantly differ from those of long-haul freight. Also, illegal parking models tend to be sample-specific, since what constitutes a violation is contingent upon local regulations and practices even though factors found to be associated with illegal parking are comparable.

Key aspects of population variation are the approach and level of classification of commercial vehicles, and this may result in considering vehicles that are not only deployed for goods delivery and service provision. This is particularly the case in the studies which analyse the determinants of illegal parking frequency by commercial vehicles based on citations databases (Kawamura et al., 2014; Kim & Wang, 2021; Wenneman et al., 2015). This challenge affects accurate identification and quantification of the effects of vehicle characteristics on the parking choices of freight vehicles. Where the travel and parking behaviours of commercial vehicles are concerned, it is typical that the decision-making process is the result of negotiation and consensus between several stakeholders. In the reviewed studies, the decision-making unit is limited to one entity, usually the commercial vehicle driver, and the common assumption is that other parties’ influences are accounted for by this single entity. Therefore, one limitation of the studies reviewed is the ambiguity of the source of parking choice decisions. Key contributors to urban last-mile goods movement include shippers, carriers, receivers, and vehicle operators. A model of parking behaviour which does not account for this complex decision-making structure may be unreliable for the evaluation of policies.

Alternatively, different nesting structures of parking decisions can be explored to not only mimic the decision-making structure in freight travel and parking, but also explore the dependencies across alternatives. For example, the choice of parking type is primarily conditional on the choice of parking location, which is in turn controlled by pre-planned vehicle tours set by logistics operators and receivers and constraints of the parking infrastructure. Similarly, parking duration is not only dependent on the activity purpose and shipment characteristics, but also on the operational features of parking facilities.

6.2. Parking cruising: an overlooked behaviour for commercial vehicles

Research shows that despite engaging in illegal parking more often than passenger vehicles, most commercial vehicles seek to park legally (Kawamura et al., 2014). Therefore, it is reasonable to hypothesise that commercial vehicles, like passenger vehicles, could engage in parking cruising (Dalla Chiara & Goodchild, 2020; Dalla Chiara et al., 2021, p. 2022). Modelling parking cruising is challenging since it is difficult to observe.

Of the reviewed studies, only Dalla Chiara and Goodchild (2020) and Dalla Chiara et al. (2022) attempt to model parking search time by delivery vehicles. Dalla Chiara and Goodchild (2020) find that parking search times are higher in areas with curb side restrictions for commercial vehicles. Higher parking turnover, as dictated by the activity time and/or time restrictions on parking space usage, also reduces parking search time. Increased capacity of designated and permissible parking for commercial vehicles reduces parking cruising. Dalla Chiara et al. (2022) simulate the impacts of real-time information on parking availability of freight loading zones, which reduces cruising. Further investigation of the parking cruising of commercial vehicles is warranted and adopting parking occupancy models for this purpose presents a future research opportunity.

6.3. Freight demand management: looking beyond parking policy

Freight demand management initiatives aim to improve the efficiency and sustainability of freight activity and operations. They comprise measures that aim to reduce the externalities of freight traffic. Despite being implemented less frequently than measures pertaining to infrastructure, traffic, and parking management, research shows that they have higher potential in terms of enhancing efficiency and sustainability (Holguin-Veras, 2016). This is because freight demand management measures directly influence freight travel and parking behaviours, targeting demand generation, rather than focusing solely on the supply-side like the parking policies evaluated in the studies included in this review. Examples of such measures include shifting delivery times to off-peak hours of traffic, consolidating shipments and/or delivery trips, coordinating delivery destinations, and modifying delivery modes (Holguín-Veras et al., 2020).

Jaller et al. (2013) show that shifts to off-peak delivery can enhance the efficiency of parking supply, such that shifts of 10% to 30% into off-peak periods generally result in considerable improvements for regular operational periods in terms of lower occupancy rates and more balanced distribution of parking demand. Campbell et al. (2018) evaluate the impacts of off-peak hour deliveries, staggered deliveries, and receiver-led consolidation on freight parking demand for a large freight traffic generator (FTG) located in Troy, New York and within Soho area in downtown New York City. Shifting a third of deliveries to off-peak hours is expected to reduce the morning-peak freight parking demand by 10% to 25%. Shifting all deliveries to off-peak hours could reduce parking demand by about 80% for the FTG in Troy and 70% for Soho.

Dalla Chiara et al. (2017) and (2020) also evaluate the impacts of goods consolidation at centralised receiving stations in freight loading and unloading bays at a large retail mall in Singapore. They find that such a policy reduces the handling time and delivery cost, improving by that the efficiency of the delivery process. Higher parking turnover and greater utilisation of loading and unloading bays are expected, and this subsequently reduces illegal parking frequency by commercial vehicles. This is in line with Holguin-Veras (2016), in which it is shown that shorter handling times at loading bays are associated with less parking cruising and illegal parking by commercial vehicles.

Delivery time, frequency, destination, and mode are key determinants of freight parking behaviour, and any measure which involves modifying one or more of these aspects of freight activity must subsequently affect the parking choices of freight vehicles. Empirical methods and analytical models allow for better identification and quantification of the effects of freight demand management measures at disaggregate levels, and hence support more informed policy implementation.

6.4. Freight data: data fusion to overcome challenges

Data is key to developing accurate models, but data collection is challenging for freight transportation. Synthesis of the reviewed studies reveals three main challenges to freight parking data: availability, completeness, and variability across stakeholders.

Due to data collection challenges and private sector restrictions on data sharing, it is typical for datasets used for freight travel and parking demand modelling to be smaller than those used for commute travel and parking demand modelling. This directly affects the accuracy of models, as well as their predictability and transferability.

Where data is available, standardisation is critical. Data standardisation entails retrofitting data from different sources to a consistent and usable format. In the reviewed studies, few constructed datasets by fusing multiple data sources; this technique resolves the sample size limitations of freight transport data. However, data fusion methods require rigorous data cleaning and standardisation to ensure semantic, temporal, and spatial heterogeneity among different datasets. Still, it is recommended that future research focusses on how big data and advanced fusion techniques can be deployed to develop robust parking choice models.

In conclusion, this scoping review is the result of a rigorous search and synthesis. Quality information sources are considered to ensure evidence reviewed is of high calibre. However, this review has two main limitations. First, studies reported in a language other than English are not considered, and so it is possible that relevant research might have been disregarded. Second, it was difficult to map out a common ground among the reviewed studies, particularly due to differences in study settings, ambiguity in the sources of parking behaviour decisions, and variability of the behavioural determinants considered. To partially overcome this, we resorted to classifying the reviewed studies not only according to analysed outcomes, but also, per used methods, considered behaviour determinants, evaluated parking policies, and deployed data sources. Despite these limitations, this review presents a comprehensive summary of the existing evidence on the parking behaviours of commercial vehicles to guide further research and practice in the field.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

We acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC); City of Toronto; Region of Peel for their funding support.