Uploaded November 2025 | Updated September 2026, 3 weeks ago
Accepted at IJRR.
Arxiv: arxiv.org/abs/2511.21312
Code: github.com/ntnu-arl/sdf-nmpc
Project page: ntnu-arl.github.io/sdf-nmpc
This paper introduces a Neural Nonlinear Model Predictive Control Framework (NMPC) framework for mapless, collision-free navigation in unknown environments with Aerial Robots, using onboard range sensing. We leverage deep neural networks to encode a single range image, capturing all the available information about the environment, into a Signed Distance Field (SDF). The proposed neural architecture consists of two cascaded networks: a convolutional encoder that compresses the input image into a low-dimensional latent vector, and a Multi-Layer Perceptron that approximates the corresponding spatial SDF. This latter network parametrizes an explicit position constraint used for collision avoidance, which is embedded in a velocity-tracking NMPC that outputs thrust and attitude commands to the robot. First, a theoretical analysis of the contributed NMPC is conducted, verifying recursive feasibility and stability properties under fixed observations. Subsequently, we evaluate the open-loop performance of the learning-based components as well as the closed-loop performance of the controller in simulations and experiments. The simulation study includes an ablation study, comparisons with two state-of-the-art local navigation methods, and an assessment of the resilience to drifting odometry. The real-world experiments are conducted in forest environments, demonstrating that the neural NMPC effectively performs collision avoidance in cluttered settings against an adversarial reference velocity input and drifting position estimates.
Accepted at IJRR.
Arxiv: arxiv.org/abs/2511.21312
Project page: ntnu-arl.github.io/sdf-nmpc
Github: github.com/ntnu-arl/sdf-nmpc
Accepted at IJRR.
Arxiv: arxiv.org/abs/2511.21312
Code: github.com/ntnu-arl/sdf-nmpc
Project page: ntnu-arl.github.io/sdf-nmpc
This paper introduces a Neural Nonlinear Model Predictive Control Framework (NMPC) framework for mapless, collision-free navigation in unknown environments with Aerial Robots, using onboard range sensing. We leverage deep neural networks to encode a single range image, capturing all the available information about the environment, into a Signed Distance Field (SDF). The proposed neural architecture consists of two cascaded networks: a convolutional encoder that compresses the input image into a low-dimensional latent vector, and a Multi-Layer Perceptron that approximates the corresponding spatial SDF. This latter network parametrizes an explicit position constraint used for collision avoidance, which is embedded in a velocity-tracking NMPC that outputs thrust and attitude commands to the robot. First, a theoretical analysis of the contributed NMPC is conducted, verifying recursive feasibility and stability properties under fixed observations. Subsequently, we evaluate the open-loop performance of the learning-based components as well as the closed-loop performance of the controller in simulations and experiments. The simulation study includes an ablation study, comparisons with two state-of-the-art local navigation methods, and an assessment of the resilience to drifting odometry. The real-world experiments are conducted in forest environments, demonstrating that the neural NMPC effectively performs collision avoidance in cluttered settings against an adversarial reference velocity input and drifting position estimates.
Accepted at IJRR.
Arxiv: arxiv.org/abs/2511.21312
Project page: ntnu-arl.github.io/sdf-nmpc
Github: github.com/ntnu-arl/sdf-nmpc
