Research map / Reinforcement learning / Visual and representation RL
Reducing AUV Energy Consumption Through Dynamic Sensor Directions Switching via Deep Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes RECS, a DQN-based method that dynamically switches AUV sensing directions and radii, guided by weighted grid edges and APF re-planning, to cut sensing energy use in simulated 2D environments.
Problem
AUVs have limited battery energy, and their sensing systems consume a large share of it, yet prior work mostly reduces computing and movement energy rather than sensing energy.
Approach
Divide the 2D map into square areas sized from the sensing radius, weight area edges by the planned path, and use DQN to choose which edges to sense with direction-specific, radius-scaled sensors; APF re-plans the path when obstacles are sensed.
Key results
- Average energy reduction of 53.4% vs full sensors on across six settings (minimum 50.8%), with the abstract reporting 53.48%.
- Average sensed edges per step below 2 (SE) and average sensing directions about 2.4 per action, showing selective sensing.
- Energy reduction stays stable at 51.2% and 53.2% for sensing radii 7.07 and 14.14, with travel distance increase under 1%.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use RECS dynamic sensor-direction switching instead of full sensors on | AUV sensing system · Evaluated environments in Table 1 | energy consumption reduction rate: 53.4% | Helped |
| Use RECS dynamic sensor-direction switching instead of full sensors on | AUV movement · Evaluated environments in Table 1 | travel distance: less than 1% | Hurt |
| Increase the sensing radius from 7.07 to 14.14 | RECS · 200×200 map | energy consumption reduction rate: 51.2% -> 53.2% | No clear effect |
Limitations
- Simulation only in static 2D environments with point-like AUV and constant speed
- Sensing energy model taken from prior settings and linearly scaled; sensing energy may exceed allocated energy
- Future work needed for 3D space and movable obstacles
Experimental setup
Datasets: simulated 100x100 and 200x200 maps with random obstacles
Benchmarks: simulated 100x100 and 200x200 maps
Metrics: energy reduction rate (ER) average sensed edges per step (SE) travel distance energy consumption average sensing directions (SD)
Baselines: full sensors on
Builds on: Deep Q-Network prioritized experience replay artificial potential field
Closest papers by meaning
- UACOF: A USV-AUV Collaboration Framework for Underwater Tasks Under Extreme Sea Conditions (Student Abstract) (AAAI 2025 · similarity 0.85)
- Power-Aware Inverse-Search Machine Learning for Low Resource Multi-Objective Unmanned Underwater Vehicle Control (Student Abstract) (AAAI 2024 · similarity 0.84)
- AoI-MDP: An AoI Optimized Markov Decision Process Dedicated in the Underwater Task (Student Abstract) (AAAI 2025 · similarity 0.82)
- Vision-Language Navigation with Energy-Based Policy (NeurIPS 2024 · similarity 0.80)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.79)
- Autonomous Goal Detection and Cessation in Reinforcement Learning: A Case Study on Source Term Estimation (AAAI 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.