As more and more robots are used to perform tasks in homes, offices and warehouses, there will be a need to scale algorithms to large fleets of robots to allow for fast, reliable and safe operation. These algorithms will often have to leverage the power of the Edge and the Cloud (together called the Fog) in unison to deliver the most efficient compute that any robot requires at any point in time. In the first part of the thesis, we introduce Robot Inference and Learning as a Service for low- latency and secure inference serving of deep models that can be deployed on robots. Unique features of RILaaS include: 1) low-latency and reliable serving with gRPC under dynamic loads by distributing queries over multiple servers on Edge and Cloud, 2) SSH based authentication coupled with SSL/TLS based encryption for security and privacy of the data, and 3) front-end REST API for sharing, monitoring and visualizing performance metrics of the available models. We report experiments to evaluate the RILaaS platform under varying loads of batch size, number of robots, and various model placement hosts on Cloud, Edge, and Fog for providing benchmark applications of object recognition and grasp planning as a service. We address the complexity of load balancing with a reinforcement learning algorithm that optimizes simulated profiles of networked robots. In the second part of the thesis we propose a sampling-based multi-query graph-based motion planner for robots that parallelizes the search process using cloud-based serverless computing (AWS Lambda). Using graph-based motion planning instead of tree-based alternatives allows for efficient reuse of a precomputed road map between tasks in the same workspace. By parallelizing the precomputation and reusing exploration, a robot executing multiple actions in the same workspace can leverage an already dense graph to create more efficient motion plans in a short amount of time. We introduce an algorithm to parallelize Probablistic Roadmaps (PRM) over serverless nodes, provide proofs of asymptotic optimality and probabilistic completeness and run a suite of experiments on the Fetch robot for a pick-and-place task to measure the provided speedup.




Download Full History