Study of Variational Inference for Flexible Distributed Probabilistic Robotics

Abstract

By combining stochastic variational inference with message passing algorithms, we show how to solve the highly complex problem of navigation and avoidance in distributed multi-robot systems in a computationally tractable manner, allowing online implementation. Subsequently, the proposed variational method lends itself to more flexible solutions than prior methodologies. Furthermore, the derived method is verified both through simulations with multiple mobile robots and a real world experiment with two mobile robots. In both cases, the robots share the operating space and need to cross each other’s paths multiple times without colliding.

Toward an idiomatic framework for cognitive robotics

“…Therefore, under Stochastic message-passing , we outline a way of combining these two approaches to overcome their weaknesses. The idea of combining message-passing with stochastic variational inference we have presented before, 26 but here we generalize the idea to generative flow graphs.…”

“…The second application example relies heavily on the stochastic message-passing approach described under Stochastic message-passing to implement a simplistic form of communication between robots. 26 In this application, N unicycle type robots have to plan low-level actions toward their goals while avoiding collisions with the other robots, given knowledge about the other robots’ expected future path, as illustrated in Figure 12 . Figure 13 shows a generative flow graph of the model derived for this problem.…”

“…Figure 13 shows a generative flow graph of the model derived for this problem. 26

Figure 12 Multi-robot navigation simulation Shown is a snapshot of a simulation with 12 robots utilizing the “Planning” idiom to plan actions toward their goal while avoiding collision with each other. Colored circles with a black border indicate the current location of the robots, solid colored lines indicate samples of their future planned path distribution, colored circles indicate their current goals, and transparent colored circles indicates their last goal.

…”

“…

Figure 13 Generative flow graph for multi-robot navigation Shown is a generative flow graph of the model derived for each of the robots in a multi-robot navigation problem. 26 represents the robots internal representation of its own state as well as the state of the other robots at time τ. represents the actions of the robot itself as well as the communicated planned actions of the other robots at time τ.

…”

“…

Figure 14 UML class diagram for “UniCycleRobotPlanning” Shown is an excerpt of the UML class diagram for the class “UniCycleRobotPlanning” used for multi-robot navigation. 26 A large part of the model for multi-robot navigation aligned with the idiom in Figure 8 and could be implemented as a child class inheriting from the abstract “Planning” class overwriting the parts of the model that did not align.

…”