Path planning for unmanned surface vehicle based on predictive artificial potential field. International Journal of Advanced Robotic Systems

1. Problem Statement

The paper addresses the challenges of path planning for high-speed Unmanned Surface Vehicles (USVs) in complex maritime environments. While traditional path planning algorithms exist, they face specific limitations when applied to high-speed, dynamic USV operations:

• Local Minima: Traditional Artificial Potential Field (APF) methods often trap the USV in local minima (e.g., concave obstacles), preventing it from reaching the goal.
• Dynamic Inconsistency: Standard APF generates paths that ignore vehicle dynamics (e.g., inertia, yaw moment), resulting in unrealistic sharp turns or vibrations that a physical USV cannot execute.
• Lack of Predictive Capability: Traditional algorithms react only when the USV is close to an obstacle (reactive), rather than anticipating obstacles in advance (proactive), leading to abrupt maneuvers.
• Static Velocity: Most APF implementations use a constant velocity, failing to optimize for energy efficiency or safety by adjusting speed based on the complexity of the environment (e.g., accelerating on open water, decelerating for turns).

2. Methodology: Predictive Artificial Potential Field (PAPF)

The authors propose a Predictive Artificial Potential Field (PAPF) method that integrates vehicle dynamics and predictive capabilities into the traditional APF framework. The methodology consists of three core modifications:

A. Angle Limit (AL)

• Concept: Recognizing that a USV is a rigid body with inertia, the turning angle cannot change instantaneously.
• Implementation: The algorithm calculates an "ideal turning angle" based on the potential field gradient. This is then passed through a saturation function constrained by the USV's maximum turning angle ($\Delta \theta_m$), which is derived from the yaw moment ($M_z$) and moment of inertia ($I_{zz}$).
• Effect: This prevents sharp, physically impossible turns and reduces path vibration, ensuring the generated trajectory respects the vehicle's dynamic constraints.

B. Velocity Adjustment (VA)

• Concept: To save time and energy, the USV should not maintain a constant speed.
• Implementation: The algorithm divides motion into three states based on the magnitude of the ideal turning angle:
  1. Straightway: Accelerate to maximum speed ($V_{max}$).
  2. Feasible Cornering: Maintain cruising speed ($V_c$).
  3. Turning-around: Decelerate to minimum speed ($V_{min}$) to ensure safety during sharp maneuvers.
• Effect: This adaptive velocity strategy reduces total sailing time and prevents collisions caused by high-speed turns.

C. Predictive Potential Field

• Concept: To solve the local minimum problem and enable proactive avoidance, a new potential field is introduced that acts before the USV enters the repulsive zone of an obstacle.
• Implementation:
  • It calculates the angular deviation ($\Delta \theta_v$) between the USV's current heading and the bisector of the tangent lines to the obstacle.
  • It combines this with the distance to the obstacle.
  • The potential function is exponential ($U_{prd} = K_{prd} \exp(-\frac{d}{d_{prd}} - |\Delta \theta_v|)$), creating a smooth, long-range repulsive force that guides the USV away from obstacles early.
• Effect: This allows the USV to "see" and detour around obstacles (including concave shapes) before the traditional repulsive field becomes active, effectively solving the local minimum trap.

Algorithm Flow:

1. Compute Attractive and Repulsive Potentials.
2. Compute Predictive Potential.
3. Combine potentials to determine the ideal turning angle.
4. Apply Angle Limit (AL) to constrain the turn.
5. Apply Velocity Adjustment (VA) to set the speed.
6. Perform Gradient Descent Update to generate the next position.

3. Key Contributions

1. Integration of Dynamics: The paper successfully integrates rigid body dynamics (yaw moment, inertia) into the APF via the Angle Limit, ensuring the path is physically executable.
2. Proactive Obstacle Avoidance: The introduction of the Predictive Potential Field shifts the paradigm from reactive to proactive avoidance, allowing the USV to navigate complex concave obstacles without getting trapped.
3. Energy and Time Optimization: The Velocity Adjustment strategy dynamically optimizes speed, balancing safety (slowing down for turns) and efficiency (accelerating on straight paths), directly addressing the energy-saving requirement.
4. Enhanced Reachability: The combined approach significantly improves the ability of the USV to reach goals in scenarios where traditional APF fails (e.g., goals located near obstacles or in concave traps).

4. Results

The authors validated the PAPF through simulations comparing it against Traditional APF (TAPF) and intermediate versions (AL-only, AL+VA).

• Turning Angle:
  • TAPF: Average max turning angle of 134.5 mrad (highly erratic).
  • PAPF: Reduced to 3.5 mrad, ensuring smooth trajectories.
• Sailing Time (Steps):
  • TAPF: 4182 steps.
  • PAPF: 2226 steps (a ~47% reduction compared to TAPF).
  • Note: While AL-only increased time slightly (due to detours), the addition of VA and Predictive Potential drastically cut the time.
• Local Minimum Resolution:
  • In simulations with crescent-shaped (concave) obstacles, TAPF failed and got trapped. PAPF successfully navigated around the obstacle by detecting it early via the predictive field.
• Reachability:
  • In a scenario with 11 different goals around an obstacle, TAPF could only reach 5 of them. PAPF successfully reached all 11 goals, demonstrating superior robustness.

5. Significance

This research provides a robust framework for high-speed autonomous marine navigation. By bridging the gap between theoretical path planning and physical vehicle dynamics, the PAPF method:

• Enhances Safety: Prevents collisions by anticipating obstacles and respecting turning limits.
• Improves Efficiency: Reduces mission time and energy consumption through intelligent speed control.
• Solves Critical Failure Modes: Specifically addresses the "local minimum" problem that has long plagued potential field methods in robotics.
• Future Applicability: The authors note that the predictive framework can be extended to moving obstacle avoidance (e.g., other ships or torpedoes) by incorporating trajectory prediction, making it highly relevant for dynamic maritime environments.