SARPA

Aim: To develop algorithms for robust, efficient, and collision-free multivehicle path planning and navigation.

Brief description
In this project, we are developing algorithms for finding and optimizing collision-free paths for autonomous dumpers and trucks operating, for example, at a mine or a construction site.

Background and motivation
There is a very strong trend towards autonomous vehicles, be it passenger cars, buses, trucks, dumpers or other vehicles. More and more autonomous functionality is being added to vehicles, and prototypes of fully autonomous cars are already available. A similar trend can be seen regarding trucks and dumpers. However, there are still many open research issues pertaining to such vehicles, and the vehicle industry thus has a strong interest in continued research in this field.

Detailed description
The SARPA project (an acronym for Safe and Robust Platform for Automated Vehicles Research) aims to develop a general platform for autonomous industrial vehicles, such as trucks, dumpers etc. The project involves several partners, for example Volvo GTT. Our part of the project is to develop algorithms for multivehicle path planning in unstructured (rapidly changing) environments, particularly for dumpers operating in a mine or a construction site.

A Volvo dumper, of the kind considered in the project.

While it is straightforward to plan the path of a single vehicle in a known, controllable environment, the problem quickly becomes more complex once the number of vehicle increases while, at the same time, unstructured environments are considered, i.e. environments that change rapidly and in unpredictable ways. The system that we are developing must be able to handle ad hoc additions (and removals) of individual vehicles, but also situations in which unexpected events occur, for example emergency stops due to obstacles (e.g. people) appearing in front of a vehicle. Moreover, just finding feasible, collision-free paths in not sufficent: The paths should also be such that the minimize an arbitrary combination of additional criteria, for example regarding fuel consumption, time delays, emissions etc.

In this project, we are using stochastic optimization methods (such a genetic algorithms) combined with various methods for path planning and decision-making in order to find and optimize (in real time) the vehicles' paths.