IEEE Aerospace and Electronic Systems Magazine - January 2021 - 44

Path Planning Using Probability Tensor Flows

ACKNOWLEDGMENTS
We state that there are no conflicts of interest and no financial disclosures that apply to this article.

REFERENCES
[1] A. Rudenko, L. Palmieri, M. Herman, K. M. Kitani,
D. M. Gavrila, and K. O. Arras, " Human motion trajectory
prediction: A survey, " Int. J. Robotics Res., vol 39, no. 8,
pp 895-935, 2020.
[2] D. P. Bertsekas, Reinforcement Learning and Optimal
Control. Belmont, MA, USA: American Scientific, 2019.
[3] M. L. Puterman, Markov Decision Processes: Discrete
Stochastic Dynamic Programming. Hoboken, NJ, USA:
Wiley, 2005.
[4] H. Kappen, V. Gomez, and M. Opper, " Optimal control as
a graphical model inference problem, " Mach. Learn.,

Figure 16.
Five agents try to reach their respective goals (agents and goals
are color coded: very light for the start and light for the goal). The
picture shows nine frames (t ¼ 1; 6; 7; 8; 9; 10; 11; 12; 18) of a
rather complex behavior, where agents may have to cross narrow
openings and avoid each other by waiting, changing paths, etc.
Various behaviors are worth to note: (a) In frames t ¼ 7; 8; 9, we
can note a conflict between the dark-blue and the orange agents.
The orange agent at t ¼ 7 is blocking the passage and the darkblue agent chooses to go right at t ¼ 8 beginning a new feasible
path. However, as soon as the passage opens, the dark-blue agent
decides to go back at t ¼ 9, because that path is better than the
one he had initiated. (b) In frame t ¼ 7, because of the blockage
caused by the orange agent, the red agent decides to take an external path, rather than go trough the obstacles. (c) In frame t ¼ 10,
the green agent obstructs the passage for the blue agent that chooses first to go below at t ¼ 11. Then as soon as the passage opens,
he goes back toward that passage at t ¼ 12, because it corresponds to a better path.

vol. 87, pp. 159-182, 2012.
[5] H. Simon, Models of Bounded Rationality. Cambridge,
MA, USA: MIT Press, 1984.
[6] P. A. Ortega and D. A. Braun, " Thermodynamics as a theory of decision-making with information-processing
costs, " Proc. Roy. Soc. A, vol. 469, no. 2153, 2013.
[7] H. Attias, " Planning by probabilistic inference, " in Proc.
9th Int. Workshop Artif. Intell. Statist., 2003.
[8] M. Toussaint, and A. Storkey, " Probabilistic inference for
solving discrete and continuous state Markov decision
processes, " in Proc. 23rd Int. Conf. Mach. Learn., 2006,
pp. 945-952.
[9] M. Toussaint, " Probabilistic inference as a model of
planned behavior, " Kunstliche Intelligenz, vol. 3, 2009.
[10] S. Levine, " Reinforcement learning and control as
probabilistic inference: Tutorial and review, " 2018,
arXiv:1805.00909v3.
[11] C. L. Buckley, C. S. Kim, S. McGregor, and A. K. Seth, " The

Application of this framework to real scenes is the
objective of some of our current research efforts. Real
scenes, specially if modeled at higher resolution, will
require larger computational resources. However, graphic
parallel processors are becoming widely available for
real-time implementations. Also, the probabilistic nature
of this framework is particularly relevant to model imperfect knowledge of the environment. The map can become
probabilistic with partial known obstacles and semantic
attributes to various areas. These attributes can also model
constraints and rewards. Furthermore, agents' dynamics
can be learned form real scenes. All these topics are
beyond the introductory nature of this article and will be
reported in depth elsewhere. Some partial results for gradually discovered obstacles are already available in a recent
conference paper [31]. Updates on our work will be posted
in timely manner on our website: https://www.mlsptlabunicampania.it/research/topics/path-modeling.html.
44

free energy principle for action and perception: A mathematica
review, " J. Math. Phychology, vol. 81, pp. 55-79, 2017.
[12] T. Parr and K. J. Friston, " Generalised free energy and active
inference: Can the future cause the past? " 2018,
bioRxiv304782.
[13] M. Baltieri and C. L. Buckley, " An active inference implementation of phototaxis, " 2017, arXiv:1707.01806.
[14] A. Zenon, S. Oleg, and G. Pezzulo, " An information-theoretic perspective on the costs of cognition, " Neuropsychologia, vol. 123, pp. 5-18, Feb. 2019.
[15] T. Parr and K. J. Friston, " The anatomy of inference: Generative models and brain structure, " Frontiers Comput.
Neuroscience, vol. 12, 2018, Art. no. 90.
[16] R. Kaplan and K. J. Friston, " Planning and navigation as
active inference, " Biol. Cybern., vol. 112, pp. 323-347, 2018.
[17] S. Nair, Y. Zhu, S. Savarese, and F.-F. Li, " Causal induc-

IEEE A&E SYSTEMS MAGAZINE

tion from visual observations for goal directed tasks, "
2019, arXiv:1910.01751v1.

JANUARY 2021
https://www.mlsptlab-unicampania.it/research/topics/path-modeling.html https://www.mlsptlab-unicampania.it/research/topics/path-modeling.html
IEEE Aerospace and Electronic Systems Magazine - January 2021

Table of Contents for the Digital Edition of IEEE Aerospace and Electronic Systems Magazine - January 2021
