A.N. Daryina1, I.V. Prokopyev2

1,2 Federal Research Center «Computer Science and Control» of the Russian Academy of Sciences (Moscow, Russia)

The article deals with the problem of optimizing the control of an autonomous vehicle in real time, taking into account static and dynamic constraints. Calculating the optimal trajectory in complex environments with static and dynamic constraints requires evaluating the entire space of possible states and finding the best solution. Previous methods of control were largely based on preliminary calculations and could not respond to dynamic events appearing in real time. The main problem in implementing this approach is that when calculating optimal control, it is necessary to know not only information about the model, but also information about the environment in real time. The aim of the work was to develop a method for optimal control of a real unmanned vehicle operating in a complex production environment.

In our work, we presented a new method of active simultaneous localization and mapping based on a model predictive integral path and conducted field experiments on a real unmanned vehicle, demonstrating new capabilities consisting in the ability to match the dynamics of a system with drive constraints and minimize projected costs, including conflicting ones on a complex polygon. Using the method of simultaneous localization and mapping, optimal control is calculated based on the cost function, coordinated with the localization system and the employment map generated in real time.

The method allows the unmanned vehicle to follow the optimal trajectory in accordance with the specified quality criterion, preventing failures in the localization system due to emissions or degeneration of information, which is very important for the safe use of the unmanned vehicle.

The method provides more reliable long-term data communication in the interface, includes loop closure, detection and verification, as well as integration with object recognition. A significant contribution of our work is actually the development of the technology of the active SLAM method to prevent failures in the localization system. A significant result is the solution of the practical problem of reactive unmanned vehicle planning at a complex landfill simulating a production environment. To do this, the global trajectory passed through a slide with complex turns, along which it was not overcome in manual unmanned vehicle-control mode without operator training. In difficult conditions, when spatial constraints greatly narrow the space of permissible displacements, the strategy of choosing a state space for optimization is more effective than sampling in the control space. The experiment showed that the passage of the slide with minimal error primarily depends on the localization system of the robot.

The method can be used in mobile robotics, in a wide class of nonlinear models, as well as for solving problems of optimal control of an unmanned vehicle with various constraints, including dynamic ones.

Daryina A.N., Prokopyev I.V. The method of active simultaneous localization and mapping based on the model predictive path integral for mobile robots. Neurocomputers. 2021. V. 23. № 6. Р. 12−23. DOI: https://doi.org/10.18127/j19998554-202106-02 (In Russian).

1. Leung C., Huang S., Dissanayake G. Active SLAM using Model Predictive Control and Attractor Based Exploration. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems. 2006. P. 5026–5031.
2. Dellaert F., Kaess V. Square Root SAM: Simultaneous Localization and Mapping via Square Root Information Smoothing. The International Journal of Robotics Research. 2006. V. 25(12). P. 1181–1203.
3. Kaess M., Johannsson H., Roberts R., Ila V., Leonard J.J., Dellaert F. iSAM2: Incremental Smoothing and Mapping Using the Bayes Tree. The International Journal of Robotics Research. 2012. V. 31. P. 217–236.
4. Thrun S., Montemerlo M. The GraphSLAM Algorithm With Applications to Large-Scale Mapping of Urban Structures. The International Journal of Robotics Research. 2005. V. 25(5-6). P. 403–430.
5. Montemerlo M., Thrun S., Koller D., Wegbreit B. FastSLAM: A factored solution to the simultaneous localization and mapping problem. Proceedings of the AAAI National Conference on Artificial Intelligence. 2002. P. 593–598.
6. Kschischang F.R., Frey B.J., Loeliger H.A. Factor Graphs and the Sum-Product Algorithm. IEEE Transactions on Information Theory. 2001. V. 47(2). P. 498–519.
7. Bajcsy R. Active Perception. Proceedings of the IEEE. 1988. V. 76(8). P. 966–1005.
8. Leung C., Huang S., Kwok N., Dissanayake G. Planning Under Uncertainty using Model Predictive Control for Information Gathering. Robotics and Autonomous Systems. 2006. V. 54(11). P. 898–910.
9. Scaramuzza D., Fraundorfer F. Visual Odometry [Tutorial]. Part I: The First 30 Years and Fundamentals. IEEE Robotics and Automation Magazine. 2011. V. 18(4). P. 80–92.
10. Williams G., Drews P., Goldfain B., Rehg J.M., Theodorou E.A. Aggressive driving with model predictive path integral control. IEEE International Conference on Robotics and Automation. 2016. P. 1433–1440.
11. Williams G., Aldrich A., Theodorou E.A. Model predictive path integral control: From theory to parallel computation. Journal of Guidance, Control, and Dynamics. 2017. P. 1–14.
12. https://github.com/AutoRally/autorally [Elektronnyy resurs]
13. Daryina A.N., Prokopiev I.V. An optimization method of the unmanned vehicle controller’s parameters based on the optimization of particles’s roy. In: Procedia Computer Science. 14th International Symposium «Intelligent Systems». 2021. P. 787–792.
14. Rosenblatt J.K. Optimal selection of uncertain actions by maximizing expected utility. Proc. IEEE Int. Symposium on Computational Intelligence in Robotics and Automation. 1999. P. 95–100.
15. Guizilini V., Ramos F. Towards real-time 3d continuous occupancy mapping using hilbert maps. The International Journal of Robotics Research. 2018. V. 37. P. 566–584.
16. Daryina A.N., Prokopiev I.V. Unmanned vehicle’s control real-time method based on neural network and selection function. In: Procedia Computer Science. 14th International Symposium «Intelligent Systems». 2021. P. 217–226.
17. Hart P.E., Nilsson N.J., Raphael B. Correction to «A Formal Basis for the Heuristic Determination of Minimum Cost Paths». SIGART Newsletter. 1972. V. 37. P. 28–29.
18. Kuwata Y., Teo J., Karaman S., Fiore G., Frazzoli E., How J.P. Motion planning in complex environments using closed-loop prediction. AIAA Guidance, Navigation and Control Conference and Exhibit, Honolulu, HI. 2008. Art. num. AIAA20087166
19. Mur-Artal R., Tardós J.D. ORB-SLAM2: an open-source SLAM system for monocular, stereo, and RGB-D cameras. IEEE Transactions on Robotics. 2017. V. 33(5). P. 1255–1262.
20. https://wiki.ros.org/navigation [Elektronnyy resurs]
21. Drews P., Williams G., Goldfain B., Theodorou E.A., Rehg J.M. Aggressive Deep Driving: Combining Convolutional Neural Networks and Model Predictive Control. 2017. arXiv:1707.05303