Supervised Autonomous Cooperation of Intelligent Specialty Vehicles

Staff Mentor: T. Keviczky (Tamas)

Other Mentor(s):

Andrea Simonetto


Distributed and large-scale systems; Distributed control; Multi-agent systems; Optimal and model predictive control


The agricultural, forestry, construction, mining and similar industries are sectors where several specialty vehicles are typically working in close vicinity in a changing and partially unknown environment to reach a collective goal. Although the individual vehicles are highly advanced, they are still operated manually and do not communicate with each other. Further improvements to the productivity and safety of these vehicles requires them to become aware of their environment and to interact in a coordinated way.

The main goal of this MSc project is to develop a framework for supervised autonomous cooperation of intelligent specialty vehicles. These vehicles will have to be able to work together within a group of varying size and accomplish a number of different tasks, including obstacle avoidance and online path planning for field area coverage. Computational requirements of centralized path planning for complex, large areas are prohibitive even for a single vehicle. Considering multiple vehicles complicates the problem, but also presents an opportunity of collaboration and distribution of computational tasks. The combinatorial optimization problem can be simplified for the price in performance (optimality) by restricting the planning horizon and resolving the planning problem after a given time interval. Important aspects of the problem are information exchange between subsystems, and cooperation between their controllers. Due to repeated online optimizations, using predictive local subsystem controllers provides numerous capabilities for both communication and cooperation.

Research will focus on algorithm and methodology development for multi-agent control systems. In particular, emphasis will be put on developing and testing cooperative path planning methods with coordination and consensus protocols in a distributed environment. The proposed approach should consider constrained dynamics, where communication between agents is limited but at the same time necessary to achieve a common objective. The developed algorithms will be tested on an experimental mobile robot testbed, which is currently under construction.

Group of harvesters

© Copyright Delft Center for Systems and Control, Delft University of Technology, 2017.