BOG and Fuzzy Controllers Based Multimodal Collision Avoidance for Industrial Manipulators

In this paper a new approach, utilizing a new paradigm built on the fusion of both Bayesian occupancy grid (BOG) and fuzzy logic controller (FLC), is presented. The aim of this work is to integrate a probabilistic approach based on the Bayesian Occupancy grids together with the fuzzy logical-based approach. The advantages of this method are several: first of all we can model the behaviours of the obstacles, instant by instant, with a probabilistic model, even if detected from different families of sensors, in order to achieve sensor fusion and robustness to uncertainty of data. On the other hand, the fuzzy logic control helps the algorithm to converge faster to the optimum value of speed override when the obstacle is distant enough, taking also into account the position of the obstacles with respect to the heading of the robot Tool Center Point (TCP). Furthermore the FLC has behavioural features, in the sense that it takes into account the behaviours of obstacles: this makes the control system more accurate. Another important aspect of the presented method is that it merges BOG environmental approach to FLC robot-centric one. This gives to the system a more complete vision, since from one side BOG has an "external" view of the scene (utilizing a map of the working area) and the FLC has an "internal" view (utilizing a robot-centric framework). Simulation results show that the robot override speed is adapted constantly to avoid collision with the obstacles, adapting its behaviours to the level of available knowledge with a smooth control law which merges the stochastic approach to the deterministic one. Extensions to control of acceleration of the TCP, integration of the algorithm with the management of robot restricted areas and to the prediction of the trajectories of obstacles are proposed as future development.