A Concurrent Constraint Programming Approach for Trajectory Determination of Autonomous Vehicles

This paper shows an approach to model trajectory determination, using concurrent constraint programming, for autonomous multi-agent systems in charge of automatically managing the mission of an uninhabited vehicle. In the approach suggested by GP01] and BGP99], the task of the constraint solver is to automatically generate a plan for autonomous vehicles according to the changes of the environment. The generation of the plan considers the constraints associated with the trajectory of the vehicle, the scheduling of the mission goals and the management of resources. Thus, the task of the solver is to nd a solution (a plan) that satisses all those constraints. We extend their approach by introducing the notion of elimination of irrelevant variables, considering a more sophisticated search strategy based on the structure of the environment, and allowing the deenition of the conditions of the environment by demand. In this article, we will focus on the issue of trajectory determination and management of contingencies, treating the determination of the trajectory as the problem of nding the best way of going from one node to another in a directed graph. Even though there are well known algorithms to deal with the standard versions of this problem (chapter 25 of CTR90]), they are no longer practical options when additional constraints over the graph are taken into account. For instance: 1. It may be requested that a particular node must be part of the solution. The same applies for edges. 2. An order among the nodes may be deened so that the order is respected in the solution found. For example, if in a given graph a node A is less than a node B the node A should be visited before the node B. For edges, the situation could be the same. 3. Each node and each edge could be associated with a type and it could be demanded that the solution must use at least (or at most) a number of nodes/edges of a given type. 4. The node and edge types could be used to order solutions. For example, it could be stated that a solution X is better than a solution Y if the number of used edges of a particular type in X is less than the number of used edges of that type in Y. ?