Using a cognitive architecture for general purpose service robot control

Figures & data

Figure 1. Structure of the overall system developed to translate high-level commands into robotic actions.

Table 1. Comparison between different cognitive architectures.

	CRAM	SS-RICS	ACT-R	SOAR
Main orientation	Robot	Robot	Cognition	Cognition
Symbolic reasoning	Simple	Simple	Complete	Complete
Low-level control	Integrated	Integrated	Middle complex	Complex implementation
Underlying structure	–	–	Complex	Easy understanding/debug
Programming complexity	Complex	Complex	Complex	Simple and elegant
Underlying architecture	ACT-R	ACT-R	–	–
Programming learnability	–	–	Easy and scalable	Complex

Note: It must be noted that ACT-R is the engine for CRAM and SS-RICS.

Table 2. Table of skills available at the robot and their associated actions.

Skill	Action
Go to	Navigate to a location
Introduce himself	Talk about himself
Follow person	Follow a specific person in front of him
Search objects	Look for objects in front of him
Search person	Look for someone in the area
Grasp object	Grasp a specific object
Deliver object	Deliver an object to the person or place in front
Memorize person	Learn a person's face and store his name
Exit apartment	Look for the nearest exit and exit the area
Recognize person	Check the person in front as already known and retrieve the person's name
Point at an object	Point the location of a specific object

Figure 2. The robot platform REEM: (a) the REEM humanoid robot used in the experiments and (b) REEM head with Kinect sensor included.

Figure 3. REEM robot at the experimental environment that mimics a home.

Figure 4. Sequence of actions done by the REEM robot to solve the command ‘Bring me a drink’: Listen to command, request missing information and acknowledge full command, go to kitchen, look for objects, detect energy drink, grasp energy drink, return to master, deliver energy drink.