|
" Motion Planning and Control of Autonomous Mobile Robots: Model Based and Model Free Methods "
Fahad, Muhammad
Guo, Yi
| Document Type
|
:
|
Latin Dissertation
|
| Language of Document
|
:
|
English
|
| Record Number
|
:
|
1052903
|
| Doc. No
|
:
|
TL52020
|
| Main Entry
|
:
|
Fahad, Muhammad
|
| Title & Author
|
:
|
Motion Planning and Control of Autonomous Mobile Robots: Model Based and Model Free Methods\ Fahad, MuhammadGuo, Yi
|
| College
|
:
|
Stevens Institute of Technology
|
| Date
|
:
|
2019
|
| Degree
|
:
|
Ph.D.
|
| student score
|
:
|
2019
|
| Note
|
:
|
193 p.
|
| Abstract
|
:
|
Robots are increasingly playing an important part in our daily lives and have touched our day to day activities in many ways. Navigation in complex environments is an important research topic in robotics. This dissertation contributes to the field of plume monitoring in a marine environment and the field of robot navigation in an environment occupied by people. Both model based and learning based planning and control methods are studied in this dissertation. In Part I, we study plume monitoring using autonomous surface vessels, and in Part II of the dissertation, we study the navigation of robots in a pedestrian occupied environment. Environmental monitoring using autonomous vessels is an important research topic. The problem of source seeking has been studied in existing work, but level curve tracing and plume front monitoring in this domain are more challenging due to the dynamic nature of pollution plume propagation. In Part I of this dissertation, we focus on this latter problem. Current research work in level curve tracing (and source seeking) in marine applications assumes simplified environments that are not representative of the plumes observed in field experiments. In general, there is a lack of plume propagation models suitable for robotic studies. This dissertation addresses this problem by performing static and dynamic surveys to study the fine-scale characteristics of marine plumes. These fine-scale characteristics are used as a baseline to propose an Eulerian advection diffusion plume model, with a time varying plume source and time varying flow field. Concentration time series of stationary and dynamic sampling points across cross-sections and plume fronts of simulated plumes using this modified model are collected and analyzed for their statistical parameters. These parameters are then compared with statistical parameters of the time series collected during experimental plume surveys. The comparison validates that the simulated plumes exhibit fine-scale qualitative and quantitative characteristics similar to experimental plumes. Furthermore, a tracking control law is proposed for an unmanned surface vessel to track the plume front. The robustness of the proposed control law is studied in simulations using both Eulerian advection diffusion plume model and probabilistic Lagrangian environmental model. Field testing of this control law to track level concentration curves is performed in a plume generated using Rhodamine dye as a pollution surrogate in a near-shore marine environment. The proposed controller performed reasonably well in field experiments, despite the challenges presented such as high concentration gradients and the perturbations caused by the motion of the USV in the plume. The increasing involvement of robots in everyday lives of people have highlighted the importance of human robot interaction, specifically, navigation in an environment occupied by humans such as offices, malls and airports. Part II of this dissertation studies this navigation problem. The human motion model consists of several complex behaviors that are difficult to capture using analytical models. Existing analytical models, such as the social force model, although commonly used, are unable to generate realistic human motion and do not fully capture behaviors exhibited by humans. These models are also dependent on various parameters that are required to be identified and customized for each new simulation environment. Artificial intelligence has received booming research interest in recent years. Solving problems that are easy for people to perform but difficult to describe formally is one of the main challenges for artificial intelligence. The human navigation problem falls directly in this category, where it is hard to define a universal set of rules to navigate in an environment with other humans and static obstacles. Reinforcement learning has been used to learn model-free navigation, but it requires a reward function that captures the behaviors intended to be inculcated in the learned navigation policy. Designing such a reward function for human like navigation is not possible due to complex nature of human navigation behaviors. We propose to use direct human trajectories to learn both the reward function and navigation policy that drives the human behavior. Using a database of real-world human trajectories collected over a duration of 90 days inside a mall, we develop a deep inverse reinforcement learning approach that learns the reward function capturing human motion behaviors. Furthermore, this dataset was visualized in a robot simulator to generate 3D sensor measurements using the LIDAR sensor onboard the robot, and a generative adversarial imitation learning based method is developed to learn the human navigation policy using the human trajectories as expert demonstration. The learned navigation policy is shown to be able to replicate human trajectories both quantitatively, for similarity in traversed trajectories, and qualitatively, in the ability to capture complex human navigation behaviors, such as leader follower behavior, collision avoidance behavior, and group behavior. In summary, this dissertation investigates both model based control and model free learning approaches for control of autonomous mobile robots and motion planning, and demonstrates their successful applications to pollution plume monitoring in a marine environment and the navigate task in pedestrian occupied environments, respectively. The marine pollution plume tracking controller was tested to perform reasonably well in field experiments. The human navigation policy learned using measured human trajectories shows performance similar to current state of the art method, using LIDAR sensors, which has not been previously achievable.
|
| Descriptor
|
:
|
Robotics
|
| Added Entry
|
:
|
Guo, Yi
|
| Added Entry
|
:
|
Stevens Institute of Technology
|
| |