| Author: | Khan, Ameer Tamoor |
| Title: | Metaheuristic-based control framework for articulated robots in smart spaces |
| Advisors: | Li, Shuai (COMP) Wu, Xiaoming (COMP) |
| Degree: | Ph.D. |
| Year: | 2022 |
| Subject: | Metaheuristics Heuristic programming Robots Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Computing |
| Pages: | xii, 183 pages : color illustrations |
| Language: | English |
| Abstract: | In recent years, the notion of intelligent robotic agents in smart spaces has emerged as a huge opportunity for growth and study. An effective framework for smart robot path-planning and control remains a hot issue among control and machine learning researchers. Our thesis focuses on developing a metaheuristic algorithm to design a control framework for articulated robots to assist humans and autonomously execute tasks assigned to implement the concept of smart spaces, such as smart homes, shopping malls, home care services, rescue operations, and hospitals. First, we will discuss our nature-inspired meta-heuristic optimization algorithm. It is inspired by the food searching nature of the beetles. The beetle registers the smell of food on its two antennae (left and right), and based on the higher intensity; it either moves in the left direction or right. The algorithm is known as BAS, and it suffers from a limitation known as "virtual particle." Due to the limitation, the algorithm becomes computationally expensive and time-consuming for complex problems. We proposed a variant of BAS known as BASZNN (Beetle Antennae Search With Zeroing Neural Network) to overcome the limitation, which is computationally efficient and consumes less time in solving real-world problems. We performed the time and space analysis and showed that BASZNN has linear complexity. We applied it to several unconstrained and constrained benchmark optimization problems to evaluate the algorithm's performance. To further test the efficiency of our algorithm, we compared it with several state-of-the-art swarm-based heuristic algorithms, i.e., PSO, GA, ALO, and FA. Second, we designed an optimization-driven problem for path planning and the obstacle avoidance of the articulated redundant robotic agent. Our formulated problem works in the forward kinematic domain instead of inverse kinematics, which is complex to formulate and evaluate. Our optimization problem allows the robotic arm to work in task space rather than joint space. Considering the complex smart-home environment, which includes static and dynamic obstacles, we based our framework on redundant manipulators to give flexibility in controller design through extra degrees of freedom. The redundant manipulator has the dexterity and agility to simultaneously perform the primary (trajectory following, lifting, and placing objects) and secondary tasks (obstacle avoidance, power efficiency, physical limit escape, and torque minimization) efficiently. Therefore, we formulated a unified optimization problem for the tracking control of the robot while avoiding obstacles. We divided the problem into sub-optimization problems, i.e., robotics path-planning and obstacle avoidance, and later unified them into a single optimization problem with the penalty term. For obstacle avoidance, we employed GJK (Gilbert-Johnson-Keerthi) algorithm since it considers the 3D geometry of the obstacles rather than considering them as point objects. Third, we considered another problem for obstacle avoidance under strict constraints. We consider the path-planning problem of the surgical robot under the remote center of motion (RCM) constraints, which is required to ensure the safety of patients. We formulated the optimization problem on a similar principle as mentioned above. We used our proposed algorithm to find the optimal trajectory that avoids the obstacles for the surgical bot and strictly obeys the RCM constraints to perform the procedure efficiently. Fourth, we extended our optimization problem to multiple mobile robotic agents. The objective function is optimized using the BASZNN algorithm. The objective function includes the two path-planning problems, i.e., path-planning for the mobile base and the path planning of a 7-DOF robotic arm mounted on a mobile base. The algorithm is tested in two scenarios: Robotic agents follow the reference trajectory individually and collaboratively. Both cases were successfully tested with BASZNN, and it was able to control multiple agents simultaneously. Fifth, we presented an optimization problem that includes human assistance by robotic agents in a simulated smart-home environment. Our formulated problem is based on three objectives, i.e., path-planning, including obstacle avoidance, lifting an object, and assisting a human. We simulated a smart-home environment (with static and dynamic obstacles) with three robotic agents (KUKA LBR IIWA 7 mounted on P3-DX) to accomplish the task. The objective is to assist humans in moving a table from one place to the other. There are three phases in performing the task. In the first phase, all three robotic agents successfully found the optimal trajectory from their location to the targeted table. In the second phase, the robotic agents lift the table along with humans under several strict constraints to maintain the table's stability while lifting it. In the third phase, the robotic agents assist humans in moving the table while maintaining a specific distance from each other and the human. Our proposed algorithm BASZNN solved the optimization problem by converging to the optimal solution; as a result, the agents performed the tasks successfully. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/12121