The rapid growth of robotics in everyday life demands solutions that enable robots to navigate unbounded and changing environments efficiently. Current methods can individually achieve spatial mapping and dynamic object detection and tracking, but integrating these two crucial abilities remains a significant challenge. A new framework, SMAT (Simultaneous Mapping and Tracking), aims to address this issue by providing a self-reinforcing mechanism for mutual improvement of mapping and tracking performance.
SMAT Framework
The SMAT framework integrates the front-end dynamic object detection and tracking module with the back-end static mapping module using a self-reinforcing mechanism. This combination promotes mutual improvement of mapping and tracking performance, enabling robots to adapt to changing environments in real-time. The system can run on a CPU-only onboard computer, making it a practical solution for real-world applications.
The SMAT framework consists of two main components:
- Dynamic Object Detection and Tracking Module: This module is responsible for detecting and tracking dynamic objects in the environment using LiDAR perception.
- Static Mapping Module: This module creates a map of the static environment using geometric information from LiDAR perception.
The self-reinforcing mechanism between these two modules enables mutual improvement of mapping and tracking performance. The SMAT framework uses a combination of techniques, including:
- LiDAR-based Perception: The framework relies on LiDAR perception to gather geometric information about the environment.
- Geometric Reasoning: The framework employs geometric reasoning to update the map and track dynamic objects in real-time.
The SMAT framework has several advantages over existing research:
Scalability
SMAT can handle large-scale urban environments, outperforming existing research in terms of map dependence and experimental site scale. This is due to its ability to adapt to changes in the environment and find alternative paths when necessary.
Flexibility
Unlike solutions that rely heavily on pre-built maps, SMAT can adapt to changes in the environment and find alternative paths when necessary. This makes it a versatile solution for urban robotics.
Privacy
The framework relies on geometric information from LiDAR perception, which better protects people’s privacy compared to image-based perception.
Resource Efficiency
SMAT does not require extensive training data or GPU computation resources, making it a plug-and-play solution for real-world deployments.
Real-World Applications
In tests, the SMAT framework demonstrated its ability to achieve successful long-range navigation and mapping in multiple urban environments using just one LiDAR, a CPU-only onboard computer, and a consumer-level GPS receiver. The system can navigate in unknown and highly varied scenarios without relying on pre-built maps or heavy computational resources, making it a promising solution for real-world applications.
Some potential applications of the SMAT framework include:
- Autonomous Delivery: SMAT can be used to enable autonomous delivery robots to navigate through complex urban environments.
- Urban Exploration: SMAT can be used to enable robots to explore and map unknown urban environments.
- Search and Rescue: SMAT can be used to enable search and rescue robots to navigate through disaster-stricken areas.
Conclusion
The SMAT framework offers a promising solution for long-range navigation in unbounded and dynamic urban environments. By enabling robots to navigate without prior knowledge of the workspace or global maps, SMAT has the potential to transform the way robots are deployed in cities and beyond.
Future research will focus on the framework’s scalability and real-world deployment in various urban environments. The SMAT framework is a significant step towards enabling robots to navigate complex and dynamic environments, paving the way for widespread adoption of robotics in everyday life.
References
Tingxiang Fan, Bowen Shen, Yinqiang Zhang, Chuye Zhang, Lei Yang, Hua Chen, Wei Zhang, and Jia Pan. SMAT: A Self-Reinforcing Framework for Simultaneous Mapping and Tracking in Unbounded Urban Environments. arXiv preprint arXiv:2304.14356 (2023).
https://arxiv.org/abs/2304.14356
This work is licensed under the Creative Commons Attribution 4.0 International License.
To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.