NSL Paper Seminar 03042020

Summary

Key Points

• The home‑care routing problem is modeled as a Time‑Window Vehicle Routing Problem (TWVRP) with multiple workers, patients, and constraints. 3:17
• Three categories of constraints are defined: continuity of care (hard), geographical clustering (soft), and time‑window availability (soft/hard). 5:03
• Penalties are weighted: hard‑constraint penalties receive a very large constant to enforce feasibility, while soft‑constraint penalties are weighted by λ1 and λ2. 17:28
• The algorithm encodes solutions as assignments of workers to tasks and uses brightness based on total travel duration plus penalty terms for constraint violations. 17:34
• Experiments compare the Firefly algorithm with Particle Swarm Optimization and a local‑search method on two benchmark instances. 18:59
• Results show the Firefly algorithm achieves the lowest total travel time and objective function value, while meeting all constraints better than the other methods. 23:16
• The authors propose a discretized version of the Firefly algorithm to assign caregivers to patients and construct feasible routes. 24:46
• The study concludes that the discretized Firefly algorithm is effective for home‑care routing and suggests future work extending the approach to other nature‑inspired algorithms and heterogeneous robot fleets. 24:50
• Future directions include applying the discretization to Particle Swarm Optimization and exploring capacity‑varying robots (ground and aerial) within the same routing framework. 25:03