Reactive path planning for low speed UAVs and mobile robots in static environments using adaptive block coordinate descent logistic regression

Unmanned aerial vehicles (UAVs) require efficient and reliable path planning mechanisms for safe navigation in obstacle-constrained environments. This study presents an Adaptive Block Coordinate Descent Logistic Regression (ABCD-LoR) model for reactive path planning in static environments using low-cost sensor inputs. The proposed approach formulates obstacle avoidance as a three-class classification problem (no turn, left turn, right turn), where navigation decisions are made sequentially based on local observations rather than pre-computed global trajectories. The model parameters are optimized using an adaptive block coordinate descent strategy, which improves convergence stability and reduces sensitivity to noisy updates compared to stochastic gradient-based methods. Performance is evaluated using statistical model fit measures, including Pseudo R2, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), log-likelihood, and likelihood ratio tests, along with classification metrics and ROC analysis. Simulation experiments are conducted in dense static environ-ments under varying speed conditions. The proposed method is compared with classical path planning approaches, including A*, Vector Field Histogram (VFH), Fuzzy Logic Controller (FLC), Dynamic Window Approach (DWA), Artificial Potential Field (APF), and ASGD-based logistic regression. Results indicate that ABCD-LoR achieves reduced path length and lower computational time while maintaining competitive predictive performance. The study is based on a simu-lated dataset generated from infrared and ultrasonic sensor configurations. While the results demonstrate the feasibility of the approach for real-time reactive nav-igation, the limited dataset size and absence of sensor noise highlight the need for further validation in real-world and dynamic environments.