Section: Research Program

Single and Multi-robot Motion-Planning

Motion-planning is a classic and large problem of navigation robotics. In Chroma it is considered in its broad definition, that is all tasks that require one or several robots to move and to interact autonomously. In this context, we aim at designing navigation models and strategies that scale up with the complexity of the robotic system, the environment, and the social rules of the human beings that operate and may interact. This research axis is divided in two complementary challenges: (a) single-robot motion-planning in human-populated environment; and (b) multi-robot motion-planning in complex environments.

Single-robot motion-planning in human-populated environment

Motion planning in dynamic and human-populated environments defines a central challenge of robotics. Unlike static or controlled environments (known environment without uncertainty) where global path planning approaches are suitable, dealing with highly dynamic and uncertain environments requires to integrate the notion of risk. This risk can be computed by methods proposed in section 3.1.3 . Then we examine how motion planning approaches can integrate this risk in the generation and selection of the paths [51] (see Figure 1 .b for illustration).

Hovever, robots are expected to share the physical space with humans. Hence, robots need to take into account the presence of humans and to behave in a socially acceptable way. Their trajectories must be safe but also predictable. Their behavior should follow social conventions, respecting proximity constraints, avoiding people interacting or joining a group engaged in conversation without disturbing. For this purpose, we propose to integrate semantic knowledge (B. Kuipers, The Spatial Semantic Hierarchy, Artificial Intelligence, Volume 119, Issues 1–2, May 2000, Pages 191-233) and psycho-social theories of human behavior (Gibson, J. (1977). The theory of affordances, in Perceiving, Acting, and Knowing. Towards an Ecological Psychology. Number eds Shaw R., Bransford J. Hoboken,NJ: John Wiley & Sons Inc.) (Hall, E. (1966). The hidden dimension. Doubleday Anchor Books.) in the navigation framework we have developped for a few years through our Risk-based navigation algorithms [32] , [51] , [56] .

We also started to examine how motion planning approaches can help to predict the motion of rational physical agents (humans), based on the hypothesis that behavior –and, eventually, interactions– can be modeled through a cost function. This led us to explore mechanisms to learn that cost function from observed human behavior such as Inverse Reinforcement Learning (e.g. Brian D. Ziebart, N. Ratliff, G. Galagher, C. Mertz, K. Peterson, J. A. Bagnell, M. Hebert, A. K. Dey and S. Srinivasa. Planning-based Prediction for Pedestrians. International conference on Intelligent Robots and Systems, 2009.) [55] . Research in Chroma will continue in this direction with the ambition of generalizing this methodology of learning human motion in order to optimize robot motions and their interactions with humans.

Multi-robot motion-planning in complex environments.

In his reference book Planning algorithms (Steven M. LaValle, Planning Algorithms, Cambridge University Press, 2006.) S. LaValle discusses the different dimensions that made the motion-planning problem complex, which are the number of robots, the obstacle region, the uncertainty of perception and action and the allowable velocities. In particular, it is emphasized that complete algorithms require at least exponential time to deal with multiple robot planning in complex environments, preventing them to be scalable in practice (p. 320). Moreover, dynamic and uncertain environments, as human-populated ones, expand this complexity. In this context, developping simulators can help to analyze this complexity and to define planning strategies, as we done in [11] [45] , as illustrated in Fig. 1 .c.

Over the past few years, our attempts to address multi-robot motion-planning are mainly due to Multi-Agent Sequential Decision Making (MA-SDM) and Swarm Robotics (SR). MA-SDM builds upon well-known decision-theoretic models (e.g., Markov decision processes and games) and related algorithms, that come with strong theoretical guarantees. However, the expressiveness of MA-SDM models has limited scalability in face of realistic multi-robot systems (Martin L. Puterman, Markov Decision Processes; Stuart Russell and Peter Norvig, Artificial Intelligence - A Modern Approach), resulting in computational overload. In contrast, SR methods which rely on local rules – generally bio-inspired – and relating to Self-Organized Systems (D. Floreano and C. Mattiussi, Bio-Inspired Artificial Intelligence - Theories, Methods, and Technologies, MIT Press, 2008.) can scale up to multiple robots and to highly dynamic environments, but with poor theoretical guarantees [50] . SR approaches are also not geared to express complex realistic tasks or point-to-point communication between robots. In Chroma, we aim at exploiting the theoretical properties of MA-SDM and the scalability of SR as a means of developing large-scale, communicating and efficient multi-robot systems. To achieve this goal, we propose to investigate two complementary methodologies.

• First, we plan to investigate incremental expansion mechanisms in anytime decision-theoretic planning, starting from local rules (from SR) to complex strategies with performance guarantees (from MA-SDM) [13] . This methodology is grounded into our research on anytime algorithms, that are guaranteed to stop at anytime while still providing a reliable solution to the original problem. It further relies on decision theoretical models and tools including: Decentralized and Partially Observable Markov Decision Processes and Games, Dynamic Programming, Distributed Reinforcement Learning and Statistical Machine Learning.

• Second, we plan to extend the SR approach by considering the integration of optimization techniques at the local level, i.e. in robot-interaction rules. The purpose is to force the system to explore solutions around the current stabilized state – potentially a local optimum – of the system. We aim at keeping scalability and self-organization properties by not compromising the decentralized nature of such systems. Introducing optimization in this way requires to measure locally the performances, which is generally possible from local perception of robots. The main optimization techniques we will consider are Local Search (Gradient Descent), Distributed Stochastic Algorithm and Reinforcement Learning. See [54] as an illustration of such an approach in a network of intersections where the traffic of autonomous vehicles is optimized.

Beyond this general challenge, Chroma aims at developing algorithms and softwares allowing to deploy, program and test multi-robot systems, including multi-vehicle systems.