Smarter Last-Mile Logistics: An Autonomous Transport Robot with Enhanced IoT Features and Smart Route Changes based on an Obstacle Avoidance

Last-mile delivery remains time- and cost-intensive, and conventional autonomous delivery robots struggle when routes are planned once and left unchanged. We present a hybrid navigation system that fuses global IoT-guided re-routing with local, sensor-driven obstacle avoidance. The global planner continuously updates a time-varying graph with congestion and risk signals (weighted by α,β,γ), while the local layer enforces collision-free motion. Using a public Kaggle dataset for traffic/obstacle traces and complementary trials in a grid-city simulator and a campus road testbed, we evaluate three strategies on the same route cohort: Static, Local-only, and the proposed Hybrid IoT. Metrics are computed per route and summarized as route-level means (uncertainty estimated via 1,000-sample bootstrap). Hybrid IoT achieves faster missions (mean delivery time 41.21→37.46 min; − 9.09% vs Static, −22.73%vs Local-only), lower energy (−18.00% Wh/km), and safer operation (near-miss −37.50%, collision probability −86.03%, yielding 97.5% collision-free runs). Path efficiency improves from 0.72 to 0.85 (+18.06%). Stratified analyses show gains increase with congestion, and a Pareto view confirms Hybrid is non-dominated more often in the time–energy plane. Ablations indicate that IoT context and explicit congestion modelling drive most of the improvement, while risk weighting chiefly benefits safety. Overall, the results demonstrate that coupling global context with local responsiveness delivers reliable, energy-aware, and scalable last-mile autonomy.