Abstract
This article introduces physics-informed neural networks (PINNs) to the field of motion planning by utilizing a PINN framework as the steering function in the kinodynamic rapidly-exploring random tree (RRT*) algorithm. The goal of this paper is to show that PINN-based methods can be used successfully for aerospace motion planning applications. We test the RRT* algorithm coupled with PINN steering, what we call PINN-RRT*, by solving spacecraft energy-optimal motion planning problems governed by the Hill–Clohessy–Wiltshire (HCW) equations of motion and nonlinear equations of relative motion (NERM), where a deputy satellite must rendezvous with a chief satellite while avoiding spherical keep-out-zones and complying with an approach corridor. The particular PINN framework we employ approximates the solution of nonlinear two-point boundary value problems (TPBVPs), which must be solved to form connections between waypoints in the RRT* tree, via the Theory of Functional Connections (TFC). TFC enables the PINN to analytically satisfy the boundary conditions (BCs) of the TPBVP. Thus, the admissible solution search space of each nonlinear TPBVP is reduced to just the trajectories that already satisfy the BCs. Using our proposed approach, each energy-optimal TPBVP solution during the run-time of the PINN-RRT* algorithm was computed in centiseconds and with an average error on the order of machine epsilon for both the HCW and NERM dynamics.
Similar content being viewed by others
Data Availibility
Data sources employed in this paper can be made available by the authors upon reasonable request.
Notes
The variable x is substituted for \(\sigma (t)\) in the LQR-RRT* subsection.
References
Balduzzi, D., Racaniere, S., Martens, J., et al.: The mechanics of n-player differentiable games. In: International Conference on Machine Learning, PMLR, pp 354–363 (2018)
Bandyopadhyay, S., Baldini, F., Foust, R., et al.: Computationally efficient motion planning algorithms for agile autonomous vehicles in cluttered environments. Tech. rep., Pasadena, CA: Jet Propulsion Laboratory, National Aeronautics and Space Administration (2017)
Bloise, N., Capello, E., Dentis, M., et al.: Obstacle avoidance with potential field applied to a rendezvous maneuver. Appl. Sci. 7(10), 1042 (2017)
Chawla, S.K., Malhotra, V.: Autonomous active space debris-removal system. In: 2019 IEEE Aerospace Conference, IEEE, pp 1–12 (2019)
Chen, Y., Lu, L., Karniadakis, G.E., et al.: Physics-informed neural networks for inverse problems in nano-optics and metamaterials. Opt. Express 28(8), 11618–11633 (2020)
Doerr, B., Linares, R.: Motion planning and control for on-orbit assembly using LQR-RRT* and nonlinear mpc. (2020). arXiv preprint arXiv:2008.02846
Donald, B., Xavier, P., Canny, J., et al.: Kinodynamic motion planning. J. ACM 40(5), 1048–1066 (1993)
Drozd, K., Furfaro, R., Mortari, D.: Constrained energy-optimal guidance in relative motion via theory of functional connections and rapidly-explored random trees. In: AAS/AIAA Astrodynamics Specialist Conference, 2019, Univelt Inc., pp. 625–644 (2020)
Drozd, K., Furfaro, R., Schiassi, E., et al.: Energy-optimal trajectory problems in relative motion solved via theory of functional connections. Acta Astronautica 182, 361–382 (2021)
Dueri, D., Mao, Y., Mian, Z., et al.: Trajectory optimization with inter-sample obstacle avoidance via successive convexification. In: 2017 IEEE 56th Annual Conference on Decision and Control (CDC), IEEE, pp. 1150–1156 (2017)
Elbanhawi, M., Simic, M.: Sampling-based robot motion planning: a review. IEEE Access 2, 56–77 (2014)
Flores-Abad, A., Ma, O., Pham, K., et al.: A review of space robotics technologies for on-orbit servicing. Progress Aerosp. Sci. 68, 1–26 (2014)
Fricke, K., Giorgiani do Nascimento, R., Viana, F.: Quadcopter soft vertical landing control with hybrid physics-informed machine learning. In: AIAA Scitech 2021 Forum, p. 1018 (2021)
Furfaro, R., Mortari, D.: Least-squares solution of a class of optimal space guidance problems via theory of connections. Acta Astronautica (2019). https://doi.org/10.1016/j.actaastro.2019.05.050
Gao, D., Luo, J., Ma, W., et al.: Online feedback motion planning for spacecraft obstacle avoidance using positively invariant sets. Adv. Space Res. 65(10), 2424–2434 (2020)
Guang-Bin, H., Qin-Yu, Z., Chee-Kheong, S.: Extreme learning machine: theory and applications. Neurocomputing 70(2006), 489–501 (2006). https://doi.org/10.1016/j.neucom.2005.12.126
Johnston, H., Schiassi, E., Furfaro, R., et al.: Fuel-efficient powered descent guidance on large planetary bodies via theory of functional connections. J. Astronaut. Sci. 67(4), 1521–1552 (2020)
Karaman, S., Frazzoli, E.: Incremental sampling-based algorithms for optimal motion planning. Robot. Sci. Syst. 104(2), 6 (2010a)
Karaman, S., Frazzoli, E.: Optimal kinodynamic motion planning using incremental sampling-based methods. In: 49th IEEE Conference on Decision and Control (CDC), IEEE, pp. 7681–7687 (2010b)
Karimpouli, S., Tahmasebi, P.: Physics informed machine learning: seismic wave equation. Geosci. Front. 11(6), 1993–2001 (2020)
Kavraki, L., Latombe, J.C.: Randomized preprocessing of configuration for fast path planning. In: Proceedings of the 1994 IEEE International Conference on Robotics and Automation, IEEE, pp. 2138–2145 (1994)
Kuwata, Y., Teo, J., Fiore, G., et al.: Real-time motion planning with applications to autonomous urban driving. IEEE Trans. Control Syst. Technol. 17(5), 1105–1118 (2009)
Lagaris, I.E., Likas, A., Fotiadis, D.I.: Artificial neural networks for solving ordinary and partial differential equations. IEEE Trans. Neural Netw. 9(5), 987–1000 (1998). https://doi.org/10.1109/72.712178
LaValle, S.M., Kuffner, J.J., Jr.: Randomized kinodynamic planning. Int. J. Robot. Res. 20(5), 378–400 (2001)
Lawden, D.F.: Optimal Trajectories for Space Navigation. Butterworths, London (1963)
Lee, J.: Physics-informed machine learning model for bias temperature instability. AIP Adv. 11(2), 025,111 (2021)
Lembeck, C.A., Prussing, J.E.: Optimal impulsive intercept with low-thrust rendezvous return. J. Guidance Control Dyn. 16(3), 426–433 (1993). https://doi.org/10.2514/3.21027
Li, Q., Yuan, J., Zhang, B., et al.: Artificial potential field based robust adaptive control for spacecraft rendezvous and docking under motion constraint. ISA Trans. 95, 173–184 (2019)
Liu, M., Hou, M., Wang, J., et al.: Solving two-dimensional linear partial differential equations based on chebyshev neural network with extreme learning machine algorithm. Eng. Comput. 38(2), 874–894 (2020)
Ma, L., Xue, J., Kawabata, K., et al.: Efficient sampling-based motion planning for on-road autonomous driving. IEEE Trans. Intell. Transp. Syst. 16(4), 1961–1976 (2015)
Mao, Z., Jagtap, A.D., Karniadakis, G.E.: Physics-informed neural networks for high-speed flows. Comput. Methods Appl. Mech. Eng. 360(112), 789 (2020)
Mertikopoulos, P., Papadimitriou, C., Piliouras, G.: Cycles in adversarial regularized learning. In: Proceedings of the Twenty-Ninth Annual ACM-SIAM Symposium on Discrete Algorithms, SIAM, pp. 2703–2717 (2018)
Miranda, D.: 2020 NASA technology taxonomy. Tech. rep, National Aeronautics and Space Administration (2020)
Misyris, G.S., Venzke, A., Chatzivasileiadis, S.: Physics-informed neural networks for power systems. In: 2020 IEEE Power & Energy Society General Meeting (PESGM), IEEE, pp. 1–5 (2020)
Mortari, D.: Least-squares solutions of linear differential equations. Mathematics 48, 5 (2017)
Mortari, D.: The theory of connections: connecting points. Mathematics 57, 5 (2017)
Mortari, D., Johnston, H., Smith, L.: High accuracy least-squares solutions of nonlinear differential equations. J. Comput. Appl. Math. 352, 293–307 (2019). https://doi.org/10.1016/j.cam.2018.12.007
Noreen, I., Khan, A., Habib, Z., et al.: Optimal path planning using RRT* based approaches: a survey and future directions. Int. J. Adv. Comput. Sci. Appl. 7(11), 97–107 (2016)
Ogundele, A.: Nonlinear dynamics and control of spacecraft relative motion. PhD thesis, Auburn University, Auburn (2017)
Perez, A., Platt, R., Konidaris, G., et al.: LQR-RRT*: Optimal sampling-based motion planning with automatically derived extension heuristics. In: 2012 IEEE International Conference on Robotics and Automation, IEEE, pp. 2537–2542 (2012)
Peters, S., Pirzkall, C., Fiedler, H., et al.: Mission concept and autonomy considerations for active debris removal. Acta Astronautica 129, 410–418 (2016)
Raissi, M., Perdikaris, P., Karniadakis, G.E.: Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys. 378, 686–707 (2019)
Reif, J.H.: Complexity of the mover’s problem and generalizations. In: 20th Annual Symposium on Foundations of Computer Science (sfcs 1979), IEEE, pp. 421–427 (1979)
Rems, F., Risse, E.A., Benninghoff, H.: Rendezvous gnc-system for autonomous orbital servicing of uncooperative targets. In: 10th International ESA Conference on Guidance, Navigation & Control Systems (29 May-2 June 2017, Salzburg, Austria) (2017)
Richards, A., Schouwenaars, T., How, J.P., et al.: Spacecraft trajectory planning with avoidance constraints using mixed-integer linear programming. J. Guidance Control Dyn. 25(4), 755–764 (2002)
Roehrl, M.A., Runkler, T.A., Brandtstetter, V., et al.: Modeling system dynamics with physics-informed neural networks based on Lagrangian mechanics. IFAC-PapersOnLine 53(2), 9195–9200 (2020)
Sahli Costabal, F., Yang, Y., Perdikaris, P., et al.: Physics-informed neural networks for cardiac activation mapping. Front. Phys. 8, 42 (2020)
Schaub, H., Junkins, J.L.: Analytical Mechanics of Space Systems, 3rd edn. AIAA Education Series, Reston, VA, (2014). https://doi.org/10.2514/4.102400
Schiassi, E., Furfaro, R., Leake, C., et al.: Extreme theory of functional connections: a fast physics-informed neural network method for solving ordinary and partial differential equations. Neurocomputing 457, 334–356 (2021)
Starek, J.A., Schmerling, E., Maher, G.D., et al.: Fast, safe, propellant-efficient spacecraft motion planning under Clohessy–Wiltshire–Hill dynamics. J. Guidance Control Dyn. 40(2), 418–438 (2017)
Wang, J.X., Wu, J., Ling, J., et al.: A comprehensive physics-informed machine learning framework for predictive turbulence modeling. (2017). arXiv preprint arXiv:1701.07102
Wang, S., Teng, Y., Perdikaris, P.: Understanding and mitigating gradient flow pathologies in physics-informed neural networks. SIAM J. Sci. Comput. 43(5), A3055–A3081 (2021)
Webb, D.J., Berg, J.V.d.: Kinodynamic rrt*: Optimal motion planning for systems with linear differential constraints. (2012). arXiv preprint arXiv:1205.5088
Weinstein, M.J., Rao, A.V.: Algorithm 984: Adigator, a toolbox for the algorithmic differentiation of mathematical functions in matlab using source transformation via operator overloading. ACM Trans. Math. Softw. (TOMS) 44(2), 1–25 (2017). https://doi.org/10.1145/3104990
Weiss, A., Danielson, C., Berntorp, K., et al.: Motion planning with invariant set trees. In: 2017 IEEE Conference on Control Technology and Applications (CCTA), IEEE, pp. 1625–1630 (2017)
Yu, J., Haol, D., Liu, H., et al.: The intelligent planning for spacecraft autonomy in on-orbit servicing mission. In: China High Resolution Earth Observation Conference, Springer, pp. 331–342 (2019)
Zappulla, R., Virgili-Llop, J., Romano, M.: Near-optimal real-time spacecraft guidance and control using harmonic potential functions and a modified RRT. In: 27th AAS/AIAA Space Flight Mechanics Meeting, pp. 3813–3831 (2017)
Zhao, X., Shirvan, K., Salko, R.K., et al.: On the prediction of critical heat flux using a physics-informed machine learning-aided framework. Appl. Thermal Eng. 164(114), 540 (2020)
Zhou, D., Yu, Z., Zhang, Y., et al.: Translational and rotational motion planning for spacecraft close proximity using sampling-based methods. Proc. Inst. Mech. Eng. Part G 233(10), 3680–3699 (2019)
Zhou, D., Zhang, Y., Li, S.: Receding horizon guidance and control using sequential convex programming for spacecraft 6-DOF close proximity. Aerosp. Sci. Technol. 87, 459–477 (2019)
Acknowledgements
The Authors would like to thank Enrico Schiassi for fruitful discussions on the distinctions and similarities of PINNs and the TFC.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: RRT* Example
An example of an iteration of the RRT* algorithm is shown by the collection of subfigures in Fig. 13. In order to describe the algorithm the example is walked through by pointing out the lines in Algorithms 1–3 where each step takes place. Once the tree is initialized, a for-loop that extends the tree for N iterations begins. Figure 13 displays the RRT* algorithm in action past the tenth iteration. At the start of an iteration a node, denoted as \(x_\text {rand}\), is first sampled from the the C-space (line 3 of Algorithm 1). Its nearest neighbor \(x_\text {nearest}\) is then found and an edge that is steered from \(x_\text {rand}\) towards \(x_\text {nearest}\) is constructed (lines 4 and 5 of Algorithm 1). The reader should be aware that the steering threshold is \(\bar{c}=2\) for the example shown in Fig. 13, i.e., \(\texttt {Cost}(\{ x_{10},x_\text {new} \})=2\). A set of vertices \(X_\text {near}\) that can reach \(x_\text {new}\) without exceeding the metric shown in Eq (5) is then computed using the Near function (line 6 of Algorithm 1). Note that the metric shown as Eq (5) is larger than \(\bar{c}=2\) in our example. This explains why many edges have a cost greater than 2. Next, A parent node of \(x_\text {new}\) is then selected by calling ChooseParent (line 7 of Algorithm 1).
The ChooseParent procedure (Algorithm 2) searches through the vertices in the set \(X_\text {near}\) and determines the vertex that reaches \(x_\text {new}\) while incurring the least amount of accumulated cost. This minimum cost incurring vertex is denoted as \(\varvec{x}_\text {min}\) and the edge that connects \(x_\text {min}\) to \(x_\text {new}\) is \(\{x_\text {min},x_\text {new}\}\). From subfigure e it is clear \(x_\text {min}=x_2\) because the accumulated cost of edges \(\{x_\text {init},x_2\}\) and \(\{x_2,x_\text {new}\}\) equals 10, which is less than the accumulated cost of any other path towards \(x_\text {new}\). After the minimum vertex is found, \(\{x_\text {min},x_\text {new}\}\) is checked for collisions with obstacles. If \(\{x_\text {min},x_\text {new}\}\) is collision-free then \(x_\text {new}\) and \(\{x_\text {min},x_\text {new}\}\) are added to the tree (lines 9 and 10 of Algorithm 1). Note that this is why we stated earlier that Fig. 13 is an example of some iteration past the tenth iteration and not definitively that it shows iteration eleven. Many other nodes sampled at previous iterations could have resulted in building edges that collided with the obstacles, which would not have been added to the tree. Thus, the number of iterations the RRT* algorithm has performed is always greater or equal to the number of nodes in the tree. RRT* then attempts to rewire the nodes in \(X_\text {near}\) (Algorithm 3). If the edge \(\{x_\text {new},x_\text {near}\}\) has less total incurred cost than the edge that connects \(x_\text {near}\) with its parent, \(\{x_\text {parent},x_\text {near}\}\), then \(\{x_\text {parent},x_\text {near}\}\) is removed from the tree and \(\{x_\text {new},x_\text {near}\}\) takes its place. Lastly, the best solution is found by connecting the goal region, an individual state in this case, with the path in the tree that accumulates the least cost (not shown in Algorithm 1).
Appendix: B NERM TPBVP
Table 8 shows the free final time energy-optimal TPBVP governed by the NERM dynamics. Time appears implicitly through the true anomaly rate \(\dot{f}\), true anomaly acceleration \(\ddot{f}\), and chief radius \(r_c\). Time can be related to the mean anomaly following Kepler’s laws of planetary motion, given the orbital elements of the chief’s orbit. With the mean anomaly at a particular time instance, the true anomaly f at that time instance can be found using the equation of center. Again, using Kepler’s laws of planetary motion, \(r_c\) can be found once f is known at a particular time instance. Then, transforming the chief orbital elements with f into state vectors in the geocentric equatorial frame (\(\varvec{r}_{\text {gef}_c}\) and \(\varvec{v}_{\text {gef}_c}\)) allows \(\dot{f}\) and \(\ddot{f}\) to be found:
Appendix C: NERM TPBVP Constrained Expressions
The position and velocity in Table 8 can be coupled together because velocity is simply just the derivative of position. Thus, only the constrained expressions for position elements are needed, and its derivative is the constrained expression for velocity elements. Since the constraints in Table 8 are identical to those in Table 2, the constrained expression for the position and velocity for the TPBVP in Table 8 are Eqs (25) and (26). Unlike, the TPBVP in Table 1, the costates cannot be coupled with the state. Hence, they need their own expression. Luckily, they are not constrained in Table 8. Thus, their constrained expressions are just their respective free functions, i.e., linear combinations of Chebyshev orthogonal polynomials:
Plugging the constrained expressions into the system ODEs shown in Table 8 and following the steps outlined in Sect. 3, then allows the TFC solution to be obtained.
Appendix D: LERM TPBVP
Table 9 shows the fixed final time energy-optimal TPBVP governed by the LERM dynamics. Solving this TPBVP with TFC can be accomplished using the constrained expressions shown and referenced in the previous Appendix subsection.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Drozd, K., Furfaro, R. & Mortari, D. Rapidly Exploring Random Trees with Physics-Informed Neural Networks for Constrained Energy-Optimal Rendezvous Problems. J Astronaut Sci 71, 9 (2024). https://doi.org/10.1007/s40295-023-00426-3
Accepted:
Published:
DOI: https://doi.org/10.1007/s40295-023-00426-3