Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Distributed Large-Scale SLAM on Cloud Servers
The goal of this project is to investigate a distributed optimization back-end for SLAM systems, deployed on multiple cloud servers, to boost the scalability of state-of-the-art SLAM to large-scale scenarios.
State-of-the-art Simultaneous Localization And Mapping (SLAM) systems have reached a high level of maturity in single-agent [1] and multi-agent [2] scenarios. However, while showing substantial accuracy and robustness in various environments, the scalability of these solutions (in terms of participating agents and size of the workspace) remains limited. With the SLAM systems accumulating more and more information during the SLAM session, the computational capacity of the PC running the SLAM algorithms puts an upper limit to the scalability of the system. The main bottleneck in this process is the optimization scheme (Global Bundle Adjustment) usually employed by SLAM systems, crucial to the accuracy of the estimate, but with cubic complexity in the input data. The goal of this project is to investigate and implement a new type of SLAM back-end, that allows to distribute the load of global optimization to several computational instances using a distributed bundle adjustment scheme [3], significantly speeding up the optimization process while not sacrificing estimation accuracy. Combining this distributed SLAM back-end with cloud computing services, providing virtually unlimited amount of computational resources, has the potential to pave the way to true scalability of contemporary SLAM solutions.
References:
[1] Qin et al.: VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator, T-RO 2018
[2] Karrer et al.: CVI-SLAM – Collaborative Visual-Inertial SLAM, RA-L 2018
[3]Zhang et al.: Distributed Very Large Scale Bundle Adjustment by Global Camera Consensus, ICCV 2017
State-of-the-art Simultaneous Localization And Mapping (SLAM) systems have reached a high level of maturity in single-agent [1] and multi-agent [2] scenarios. However, while showing substantial accuracy and robustness in various environments, the scalability of these solutions (in terms of participating agents and size of the workspace) remains limited. With the SLAM systems accumulating more and more information during the SLAM session, the computational capacity of the PC running the SLAM algorithms puts an upper limit to the scalability of the system. The main bottleneck in this process is the optimization scheme (Global Bundle Adjustment) usually employed by SLAM systems, crucial to the accuracy of the estimate, but with cubic complexity in the input data. The goal of this project is to investigate and implement a new type of SLAM back-end, that allows to distribute the load of global optimization to several computational instances using a distributed bundle adjustment scheme [3], significantly speeding up the optimization process while not sacrificing estimation accuracy. Combining this distributed SLAM back-end with cloud computing services, providing virtually unlimited amount of computational resources, has the potential to pave the way to true scalability of contemporary SLAM solutions.
References:
[1] Qin et al.: VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator, T-RO 2018
[3]Zhang et al.: Distributed Very Large Scale Bundle Adjustment by Global Camera Consensus, ICCV 2017
Not specified
- Excellent mathematical skills (especially in non-linear optimization, computer vision knowledge is a plus)
- High degree of autonomy, high self-motivation, research-driven attitude
- Programming experience (ideally C++)
- Excellent mathematical skills (especially in non-linear optimization, computer vision knowledge is a plus) - High degree of autonomy, high self-motivation, research-driven attitude - Programming experience (ideally C++)