Signal strength assisted robot navigation in a sensor network field

Robot Navigation in unknown environments faces fundamental challenges in: (i) identification of goal locations, (ii) trajectory planning, and (iii) trajectory execution. Several research studies have explored solutions to this problem, either with technology, such as global positioning systems, or with sophisticated planning algorithms with high computational complexity. This dissertation proposes a distributed three-tiered approach to solving the problem, involving: (i) a Novel scheme for identifying goal locations using a deployed static wireless sensor network (WSN), (ii) a Probabilistic mechanism for the autonomous mobile robots (AMRs) to obtain navigation information from the WSN, and (iii) a Neighborhood way-point computation scheme for improved AMR navigation. The goal is to have AMRs navigate from any point within a WSN-covered region to an identified target via interaction with the WSN only. The constraint is that neither the AMR, nor the WSN possess global positioning information. Identification of goal locations is achieved by generating a Pseudo-Gradient in the WSN field, that has its peak closest to the target in the region. The pseudo-gradient utilizes the important artifact of WSNs, that is, Received Signal Strength (RSS). RSS displays a natural gradient with respect to distance in its distribution. The pseudo-gradient therefore mimics a gradient from any natural occurring phenomenon such as chemical leaks or heat dissipation. The trajectory planning problem is solved by having the AMR follow the pseudo-gradient in one of two ways: (i) the WSN nodes guide the AMR to the goal location through simple WSN node-to-AMR interaction, or (ii) the AMR computes the way-points by estimating the topographical distribution of the pseudo-gradient in its neighborhood. Implicit Surface Interpolation and Artificial Potential Field mechanisms are used to estimate the distribution. In this method, there is no global coordinate reference available for the region, i.e., the AMR only utilizes the relative neighborhood information and the topology of the network to reach the goal location. The challenge of trajectory execution is met by the AMR applying probabilistic mechanisms to the data gathered from the WSN. By applying a particle filtering based bearing estimation approach, the AMR recursively improves the bearing estimates of its WSN neighbor nodes. The AMR then utilizes this information, along with data from the trajectory planning tier, to execute the trajectory way-points. The performance of the proposed three-tier approach is borne out through simulation and hardware experiments. Thus, the novel contributions of this dissertation are the development of: (i) A distributed WSN-based Pseudo-Gradient algorithm, (ii) Novel neighborhood way-point computation mechanisms, and (iii) A Particle Filtering setup for probabilistic WSN neighbor-node bearing estimation.