Constraint Based Probabilistic Determination of Object State for Autonomous Robots

This paper presents a probabilistic determination algorithm for dynamic objects. Gaussian model is used along with path balancing. The predictive system uses state estimate and covariance of the tracking system and map of the environment to compute the probability distribution of the future obstacle state over a specified area. Here the system is assumed to be nonlinear. The core of this approach is a weighing matrix that balances the contribution of each vector constraint used for prediction of object state. The constraints are considered as additional feature which is injected into control law. The computed object state is integrated with navigational behaviours to enable a robot to reach its navigational goal, avoiding any hazards. All the posture is represented by polar coordinates and the dynamic equation is feedback-linearized. After determining the object state a novel sliding mode control law is used for asymptotically stabilizing the mobile robot to a trajectory so it can navigate by avoiding any hazards.