Transformers have become the powerhouse of natural language processing and recently found use in computer vision tasks. Their effective use of attention can be used in other contexts as well, and in this paper, we propose a transformer-based approach
for efficiently solving the complex motion planning problems. Traditional neural network-based motion planning uses convolutional networks to encode the planning space, but these methods are limited to fixed map sizes, which is often not realistic in the real-world. Our approach first identifies regions on the map using transformers to provide attention to map areas likely to include the best path, and then applies local planners to generate the final collision-free path. We validate our method on a variety of randomly generated environments with different map sizes, demonstrating reduction in planning complexity and achieving comparable accuracy to traditional planners.
Kinodynamic Motion Planning (KMP) is to find a robot motion subject to concurrent kinematics and dynamics constraints. To date, quite a few methods solve KMP problems and those that exist struggle to find near-optimal solutions and exhibit high compu
tational complexity as the planning space dimensionality increases. To address these challenges, we present a scalable, imitation learning-based, Model-Predictive Motion Planning Networks framework that quickly finds near-optimal path solutions with worst-case theoretical guarantees under kinodynamic constraints for practical underactuated systems. Our framework introduces two algorithms built on a neural generator, discriminator, and a parallelizable Model Predictive Controller (MPC). The generator outputs various informed states towards the given target, and the discriminator selects the best possible subset from them for the extension. The MPC locally connects the selected informed states while satisfying the given constraints leading to feasible, near-optimal solutions. We evaluate our algorithms on a range of cluttered, kinodynamically constrained, and underactuated planning problems with results indicating significant improvements in computation times, path qualities, and success rates over existing methods.
We consider the Lorentz mirror model and the Manhattan model on the even-width cylinder $mathbb{Z} times (mathbb{Z}/2nmathbb{Z}) ={(x,y):x,yin mathbb{Z}, 1leq yleq 2n}$. For both models, we show that for large enough $n$, with high probability, any t
rajectory of light starting from the section $x=0$ is contained in the region $|x|leq O(n^{10})$.
Reliable real-time planning for robots is essential in todays rapidly expanding automated ecosystem. In such environments, traditional methods that plan by relaxing constraints become unreliable or slow-down for kinematically constrained robots. This
paper describes the algorithm Dynamic Motion Planning Networks (Dynamic MPNet), an extension to Motion Planning Networks, for non-holonomic robots that address the challenge of real-time motion planning using a neural planning approach. We propose modifications to the training and planning networks that make it possible for real-time planning while improving the data efficiency of training and trained models generalizability. We evaluate our model in simulation for planning tasks for a non-holonomic robot. We also demonstrate experimental results for an indoor navigation task using a Dubins car.
We consider the periodic Manhattan lattice with alternating orientations going north-south and east-west. Place obstructions on vertices independently with probability $0<p<1$. A particle is moving on the edges with unit speed following the orientati
on of the lattice and it will turn only when encountering an obstruction. The problem is that for which value of $p$ is the trajectory of the particle closed almost surely. We prove this for $p>frac{1}{2}-varepsilon$ with some $varepsilon>0$.
Following the discovery of superconductivity in an iron-based arsenide LaO1-xFxFeAs with a superconducting transition temperature (Tc) of 26 K[1], Tc was pushed up surprisingly to above 40 K by either applying pressure[2] or replacing La with Sm[3],
Ce[4], Nd[5] and Pr[6]. The maximum Tc has climbed to 55 K, observed in SmO1-xFxFeAs[7, 8] and SmFeAsO1-x[9]. The value of Tc was found to increase with decreasing lattice parameters in LnFeAsO1-xFx (Ln stands for the lanthanide elements) at an apparently optimal doping level. However, the F- doping in GdFeAsO is particularly difficult[10,11] due to the lattice mismatch between the Gd2O2 layers and Fe2As2 layers. Here we report observation of superconductivity with Tc as high as 56 K by the Th4+ substitution for Gd3+ in GdFeAsO. The incorporation of relatively large Th4+ ions relaxes the lattice mismatch, hence induces the high temperature superconductivity.
