Kamak Ebadi

Date of Award


Document Type



Santa Clara : Santa Clara University, 2020.

Degree Name

Doctor of Philosophy (PhD)


Electrical and Computer Engineering

First Advisor

Sally L. Wood


Enabling fully autonomous robots capable of navigating and exploring unknown and complex environments has been at the core of robotics research for several decades. Mobile robots rely on a model of the environment for functions like manipulation, collision avoidance and path planning. In GPS-denied and unknown environments where a prior map of the environment is not available, robots need to rely on the onboard sensing to obtain locally accurate maps to operate in their local environment. A global map of an unknown environment can be constructed from fusion of local maps of temporally or spatially distributed mobile robots in the environment.

Loop closure detection, the ability to assert that a robot has returned to a previously visited location, is crucial for consistent mapping as it reduces the drift caused by error accumulation in the estimated robot trajectory. Moreover, in multi-robot systems, loop closure detection enables finding the correspondences between the local maps obtained by individual robots and merging them into a consistent global map of the environment. In ambiguous and perceptually-degraded environments, robust detection of intra- and inter-robot loop closures is especially challenging. This is due to poor illumination or lack-thereof, self-similarity, and sparsity of distinctive perceptual landmarks and features sufficient for establishing global position. Overcoming these challenges enables a wide range of terrestrial and planetary applications, ranging from search and rescue, and disaster relief in hostile environments, to robotic exploration of lunar and Martian surfaces, caves and lava tubes that are of particular interest as they can provide potential habitats for future manned space missions.

In this dissertation, methods and metrics are developed for resolving location ambiguities to significantly improve loop closures in perceptually-degraded environments with sparse or undifferentiated features. The first contribution of this dissertation is development of a degeneracy-aware SLAM front-end capable of determining the level of geometric degeneracy in an unknown environment based on computing the Hessian associated with the computed optimal transformation from lidar scan matching. Using this crucial capability, featureless areas that could lead to data association ambiguity and spurious loop closures are determined and excluded from the search for loop closures. This significantly improves the quality and accuracy of localization and mapping, because the search space for loop closures can be expanded as needed to account for drift while decreasing rather than increasing the probability of false loop closure detections.

The second contribution of this dissertation is development of a drift-resilient loop closure detection method that relies on the 2D semantic and 3D geometric features extracted from lidar point cloud data to enable detection of loop closures with increased robustness and accuracy as compared to traditional geometric methods. The proposed method achieves higher performance by exploiting the spatial configuration of the local scenes embedded in 2D occupancy grid maps commonly used in robot navigation, to search for putative loop closures in a pre-matching step before using a geometric verification. The third contribution of this dissertation is an extensive evaluation and analysis of performance and comparison with the state-of-the-art methods in simulation and in real-world, including six challenging underground mines across the United States.