Research Publications

#

Publications and research contributions in robotics, control systems, artificial intelligence, and machine learning.

Submitted Journal Papers

#

Paper 1: Distributed Model Predictive Control for Multi-UAV Formation with Consensus

#

Abstract: This paper presents a comprehensive implementation of distributed Model Predictive Control (MPC) for multi-UAV quadrotor formation control with consensus-based coordination and geometric SO(3) attitude control.

Key Contributions:

• Local MPC solver with formation constraints
• Consensus protocol supporting ring/mesh/star topology
• Geometric SO(3) controller using quaternion representation
• Formation planner (grid, line, circle, wedge)

Technical Details:

• Quadrotor mass: 1.0 kg
• Inertia: diag([0.01, 0.01, 0.02]) kg·m²
• PD controller: kp=0.4, kd=0.8
• Velocity limits: ±3.0 m/s

Results:

• Benchmark: 7/7 scenarios passed (100% success rate)
• Stress tests: 20/21 tests passed (95.2% pass rate)

Paper 2: Modernized Bees Algorithm for Dynamic Path Planning in Robotics

#

Abstract: This work presents a modernized implementation of the Bees Algorithm for robot path planning in dynamic environments with constraint handling.

Key Contributions:

• Adaptive parameter tuning based on convergence state
• Multi-objective optimization (path length, safety, smoothness, energy)
• Dynamic obstacle handling
• Comprehensive evaluation framework with 30-run statistical analysis
• ROS/Gazebo integration

Technical Details:

• Scout bees: 50
• Elite sites: 5, Best sites: 20
• Neighborhood size: 0.1
• Max iterations: 500

Results:

• 100% success rate across 6 scenarios
• Average planning time: 0.35 seconds
• GitHub: swarm-path-planning-bees

Paper 3: Hybrid Inverse Kinematics Ensemble with Learned Uncertainty Estimation

#

Abstract: This paper presents a hybrid ensemble architecture combining Damped Least Squares with neural network solver, along with learned uncertainty-based weighting that adapts solver contributions based on confidence.

Key Contributions:

• Hybrid ensemble combining multiple IK solver strategies
• Learned uncertainty-based weighting
• Comprehensive benchmark across 7 scenarios
• International collaboration (Italy)

Technical Details:

• Robot Model: 6-DOF UR5-like manipulator
• Damping factor: λ = 0.01
• Maximum iterations: 100
• Neural Network: MLP [3 input, 128, 128, 64, 6 output]
• Training: 4640 samples, 100 epochs, learning rate 0.005
• Ensemble weights: DLS=0.45, NN=0.55 after convergence

Results:

• 100% success rate on random targets
• 86.7% overall success rate
• Average solve time: 5ms (random targets), 8ms (overall)

Research Areas

#

Research Infrastructure

#

• Simulation: Python/NumPy, ROS/Gazebo, MATLAB/Simulink
• Testing: Custom stress test frameworks, pytest
• Documentation: LaTeX, Markdown, Jupyter Notebooks
• Version Control: Git, GitHub

Collaboration

#

International Collaborations

#

• Luca Ricci — University of Tuscia, Italy (Co-author on IK paper)

Academic Platforms

#

• IEEE ScholarOne
• Elsevier Editorial Manager
• arXiv

Research Philosophy

#

“Bridging academic theory with practical engineering applications through rigorous research and open-source contributions.”

Contact for Research Collaboration

#

For research collaboration opportunities, please contact:

• Email: molhamfetneh@gmail.com
• GitHub: github.com/molhamfetnah