Abstract
The present work provides a new approach to solve the well-known multi-robot co-operative box pushing problem as a multi objective optimization problem using modified Multi-objective Particle Swarm Optimization. The method proposed here allows both turning and translation of the box, during shift to a desired goal position. We have employed local planning scheme to determine the magnitude of the forces applied by the two mobile robots perpendicularly at specific locations on the box to align and translate it in each distinct step of motion of the box, for minimization of both time and energy. Finally the results are compared with the results obtained by solving the same problem using Non-dominated Sorting Genetic Algorithm-II (NSGA-II). The proposed scheme is found to give better results compared to NSGA-II.
Introduction
The "Box-Pushing Problem" is a fundamental benchmark in robotics research, requiring multiple agents to cooperate in moving a bulky object from an initial configuration to a target goal. This task is inherently complex due to the kinematics involved; robots must coordinate their actions to perform both translation and rotation while navigating around obstacles.
A primary challenge in this domain is the multi-objective nature of the problem. Specifically, there is a fundamental trade-off between Time Minimization and Energy Minimization. Reducing completion time typically requires applying maximum force to move the box quickly, whereas minimizing energy consumption necessitates applying minimal force to conserve battery life. Because these objectives conflict, the goal is to find a set of optimal trade-offs, known as a Pareto front, rather than a single solution.
Methodology
To address this trade-off, the researchers developed a Modified Multi-objective Particle Swarm Optimization (MOPSO) technique. Unlike standard PSO, which seeks a single global optimum, MOPSO is designed to maintain a diverse range of non-dominated solutions.
Kinematic Modeling and Local Planning
The approach relies on a precise kinematic model that accounts for the box's center of gravity and the forces applied by two mobile robots. The robots apply forces perpendicularly at specific locations to achieve either pure translation or rotation (where one robot pulls while the other pushes).
Algorithmic Enhancements
Several key components were integrated into the MOPSO framework to improve performance:
- Pareto Dominance and Grid-Based Selection: The algorithm maintains a repository of non-dominated solutions. To ensure diversity across the objective space, a grid system is used to favor particles in less-populated regions.
- Dynamic Mutation Operator: A critical addition is a nonlinear, exponentially decaying mutation rate. This allows particles to explore new areas of the search space early in the simulation to avoid premature convergence, while refining paths as the simulation progresses.
- Penalty Functions: To ensure safety, the energy objective function includes a penalty that increases significantly if the box approaches obstacles or walls, effectively guiding the search away from collisions.
- Secondary Objective Strategy: The distance to the final goal is incorporated into the objective functions as a heuristic, ensuring the robots maintain progress toward the target even when immediate paths are complex.
Experimental Evaluation
The algorithm was evaluated in two custom-designed simulated environments. World Map 1 represented a moderately complex arena, while World Map 2 featured a more complex, winding path designed to test the algorithm's ability to navigate tight constraints.
Results and Discussion
The modified MOPSO was compared against the Non-dominated Sorting Genetic Algorithm-II (NSGA-II), a standard benchmark in multi-objective optimization. Performance was averaged over 10 simulation runs to ensure statistical reliability.
In World Map 1, MOPSO achieved a total energy consumption of 33.302 KJ and a completion time of 649.51s over 10 steps. In contrast, NSGA-II required 38.115 KJ and 696.57s over 11 steps. MOPSO not only optimized the objectives better but also found a more direct path.
In the more challenging World Map 2, both algorithms required 13 steps. However, MOPSO remained superior, consuming 52.886 KJ in 810.36s, compared to NSGA-II's 56.235 KJ and 857.94s. These results demonstrate that the modified MOPSO consistently finds more efficient and faster paths than NSGA-II in complex environments.
Conclusion
This study demonstrates that a modified MOPSO approach can effectively solve the cooperative box-pushing problem by balancing time and energy efficiency. While the current model assumes static environments and perfect sensor data, future work will focus on "Dynamic MOPSO" to handle real-time environmental changes and the inherent noise of real-world robotic sensors.
Summary updated