TL;DR
This paper introduces a novel hardware-in-the-loop (HIL) framework for optimizing robotic wireless energy transfer (WET) in dynamic, obstacle-filled environments. By iteratively updating models based on real-world experiments and simulations, their approach significantly reduces charging mission times while ensuring energy delivery and avoiding collisions.
What Problem Does This Paper Address?
Powering the massive number of Internet-of-Things (IoT) devices in smart cities is a significant challenge. While wireless energy transfer (WET) offers a promising solution, its short-range nature necessitates mobile energy transmitters. Robotic WET, where a robot carries the WET transmitter, is an emerging paradigm. However, planning efficient charging strategies for these robots in unknown and dynamic environments, fraught with unpredictable obstacles, is difficult. Existing methods often treat global planning (where to charge) and local planning (how to navigate safely) separately, leading to suboptimal performance. This paper aims to bridge this gap by proposing an integrated framework that accounts for both global optimization and real-time dynamic adjustments.
Methods
The core of this paper is a Hardware-in-the-Loop (HIL) Joint Optimization Framework. This framework operates iteratively:
- Offline Planning Stage:
- Model Updates: Mathematical models (robot motion, wireless channel, energy harvesting) are updated and re-optimized using data from the previous experimental round. This allows the system to learn from real-world performance.
- Global Planning: This stage involves:
- Anchor Point Generation: Instead of traditional k-means clustering, the paper uses K-Chebychev DBSCAN to identify optimal charging locations (anchor points) that minimize the distance to the farthest energy harvester (EH) within a cluster, considering beamforming constraints.
- Anchor Point Selection: A new module prunes ineffective anchor points and selects routes that avoid dense traffic.
- Route Planning & Resource Allocation: The robot's path, charging time at each anchor, and energy beam direction are optimized jointly.
- Online Validation Stage:
- The robot follows the planned route.
- Sensing & Local Planning: Onboard sensors detect obstacles and their movements. This information is fed to a local planner that adjusts the robot's velocity in real-time to avoid collisions while trying to adhere to the global plan. This stage generates new data for the next offline planning iteration.
The system leverages a digital-twin cyber-physical platform, combining a high-fidelity Gazebo simulator with a real-world multi-robot testbed for comprehensive validation.
Datasets
Not discussed in the paper. The paper relies on simulated data from Gazebo and experimental data collected from a real-world multi-robot testbed.
Key Insights and Ideas
- The Power of Iterative Re-optimization: The HIL framework's ability to learn from experimental data and continuously refine its models is crucial. This allows the system to adapt to the nuances of real-world environments that are difficult to perfectly model beforehand.
- Anchor Point Optimization Matters: Moving beyond simple clustering, the K-Chebychev approach for anchor point selection, which considers the farthest EH and beamforming, is key to efficient energy delivery.
- Joint Optimization is Essential: Optimizing anchor points, routes, charging times, and beam directions simultaneously, rather than sequentially, leads to significant improvements in mission completion time.
- Sim-to-Real Gap is Real, but Manageable: The study highlights that while simulators like Gazebo are valuable, there's a performance difference between simulation and the real world due to factors like distributed control and sensor inaccuracies. The HIL framework helps bridge this gap.
- Dynamic Environments Demand Dynamic Solutions: The framework's success in dynamic environments with moving obstacles underscores the need for integrated global and local planning.
Key Results
- The proposed joint optimization framework significantly reduces WET mission completion time. In a perfect case scenario, it achieved a completion time of 184.3 seconds.
- In a Gazebo simulation with 5 non-cooperative robots, the HIL re-optimization resulted in a 147-second mission completion time (before HIL re-optimization) and a smaller, triangular route after HIL re-optimization, indicating improved efficiency.
- The real-world experiment showed a completion time of 150 seconds, demonstrating the effectiveness of the digital-twin approach.
- The framework successfully balances energy harvesting requirements and collision avoidance constraints.
- The paper demonstrates that increasing the number of robots in the environment leads to an increase in mission completion time, highlighting the impact of traffic conditions.
Limitations and Future Work
- Fixed Transmitter Power: The paper notes that a fixed energy transmitter with limited power (3 Watts) can struggle to deliver sufficient power to remote EHs, sometimes falling below the sensitivity threshold.
- Anchor Point Selection Impact: The study suggests that if the robot doesn't perform anchor point selection and visits all anchors, the total time increases due to the trade-off between moving and charging.
- UAV WET Comparison: While UAV WET is faster in ideal conditions, it has a significantly higher energy consumption compared to the proposed robotic WET system.
- Future Work: The authors suggest that for absolutely unknown environments, creating a global map is time-consuming. They propose Integrated Sensing and Communication (ISAC) as a promising technique to accelerate map merging by allowing robots to build and share local maps simultaneously. They also hint at the potential of multi-robot collaboration to enhance system performance.
Summary updated