Poonganam Sri Sai Naga Jyotish received his MS in Electronics and Communication Engineering (ECE). His research work was supervised by Prof. Madhava Krishna K. Here’s a summary of Poonganam Sri Sai Naga Jyotish’s MS thesis, Probabilistic Inverse Velocity Obstacles for Navigation under Uncertainty as explained by him:
Autonomous navigation gained a lot of attention in the past decade. Motion planning and collision avoidance are the core things that drive these systems. Several advances resulted in different algorithms to tackle the problem of motion planning and collision avoidance. The crucial question at this stage is how reliable and robust are these systems in a real-life scenario. Recent works in the field have proposed algorithms that account for the uncertainties that arise in a real-life scenario. But the work is sparse. In an attempt to address the reliability and robustness of the navigation frameworks in real-life scenarios, we present two approaches that incorporate the uncertainty in the sensor measurements.
First, we present a clever way, Inverse Velocity Obstacle (IVO), to achieve better collision detection based on the popularly used velocity obstacles. The proposed method stems from the concept of velocity obstacle and can be applied for both single-agent and multi-agent systems. It focuses on computing collision-free maneuvers without any knowledge or assumption on the position and the velocity of the robot. We achieve this by reformulating the velocity obstacle in an ego-centric frame. This is a significant step towards improving real-time implementations of collision avoidance in dynamic environments as there is no dependency on state estimation techniques to infer the robot’s position and velocity. We evaluate IVO for both single-agent and multi-agent cases in different scenarios and show its efficacy over the existing formulations. We also show the real-time scalability of the proposed methodology.
Next, we present a probabilistic variant of the same to handle the state estimation and motion uncertainties that arise due to the other participants of the environment. We present an algorithmic framework that computes collision-free velocities for the robot dynamic and uncertain environments. We make no assumptions on the nature of the uncertainties and model them as non-parametric probability distributions. In our Prob-abilistic Inverse Velocity Obstacle (PIVO), we pose the collision-free navigation problemas an optimization problem by reformulating the velocity conditions of IVO as chance constraints that take the uncertainty into account. We also define a lower bound on the confidence of collision avoidance maneuvers that are a result of the optimization. We demonstrate the algorithm’s ability to generate safe trajectories under highly uncertain environments.