A supervisory control policy over an acoustic communication network

Abstract

This paper presents a supervisory multi-agent control policy over an acoustic communication network subject to imperfections (packet dropout and transmission delay) for localisation of an underwater flow source (e.g., source of chemical pollution, fresh water, etc.) with an unknown location at the bottom of the ocean. A two-loop control policy combined with a coding strategy for reliable communication is presented to perform the above task. A simulator is developed and used to evaluate the trade-offs between quality of communication, transmission delay and control for a fleet of autonomous underwater vehicles supervised over a noisy acoustic communication network by an autonomous surface vessel. It is illustrated that without compensation of the effects of severe random packet dropout, localisation of an unknown underwater flow source is not possible for the condition simulated just by implementing a two-loop control policy. But a two-loop control policy combined with a strategy for reliable communication locates the unknown location of flow source.

Keywords:

• networked control system
• multi-agent system
• supervisory control

Acknowledgements

The authors thank Lara Brinon Arranz, Alexandre Seuret, Sebastien Varrier, Alain Kibangou, Julien Minet, and colleagues from Ifremer (Jan Opderbecke and Romain Piasco) for many helpful technical discussions. The authors also thank their industrial partner Ifremer (French institute for ocean research) for information on its AsterX autonomous underwater vehicles.

Notes

1. French institute for ocean research. This institute performs underwater operations for ocean science purposes.

2. For each AUV, a local control loop implements a feedback linearisation technique that replaces the vehicle nonlinear dynamics with a pseudo linear dynamics. Then, H_∞ controllers are used to compensate the effects of model inaccuracy and perturbations, including marine perturbations, and force the vehicle parameters ${\dot{r}}_i$ and $\dot{{\phi }_i}$ to follow the desired set points v_ie(φ_i) and u_i, respectively (Varrier, 2010).

3. That is, AUVs are uniformly distributed along a circumference by $\frac{2\pi }{n}$ radians each.

4. The speed of underwater sound waves is typically v_s =1500 m/s, and therefore, if the distance between AUVi and the ASV is d_i, then T_di = d_i/v_s second.

5. To exchange information (measurements, control commands, etc.) between vehicles with negligible distortion, long packets of data are used. Note that distortion is due to the quantisation of information.

6. Currently, TDMA scheme is the only commercially available technique for underwater multiple accessing without collision.

7. When the formation is perfect, i.e., it is circular, d_v = 2R₀. But, as the formation may be disturbed during search mission due to perturbations, etc., d_v is chosen to be 3R₀.

8. To model a typical underwater search mission, the parameters used in simulations, such as the description for concentration flow intensity, v₀, R₀, etc., were chosen following several technical discussions with our colleagues from Ifremer who have many years of practical experience in operational exploration with AUVs.

9. For this case, after the centre of formation enters into the ball with centre c* and radius r₀, the centre of formation shortly exits the ball.

10. For this case, after the centre of formation enters into the ball with centre c* and radius r₀, the centre of formation shortly exits the ball.

11. For each M, the empirical value of mission time (empirical T_total) is the average of the mission time obtained from simulations.

Additional information

Funding

This work is supported by the FeedNetBack European project [grant agreement number 223866].