Autonomous Smart walkers are assistive devices that promote locomotion assistance and social interaction for people with lower limb impairments and provide a safe navigation. The Probabilistic Foam method (PFM) is a sampling-based path planner that uses structures called bubbles to compute obstacle-free paths with high clearance. These bubbles propagate through the free space and ensure safe regions, meeting the safety requirements in maneuvering. The Goal-biased Probabilistic Foam (GBPF) is a variant of the PFM that improves the propagation strategy to obtain shorter paths in reduced time. In this paper, we propose a new variant of the GBPF, called Improved Goal-biased probabilistic Foam (IGBPF), to improve the planning execution time. In addition, we present the modelling of a new bubble for a smart walker, using information from the workspace. Some simulated experiments were performed using PFM, GBPF, and IGBPF to plan safe paths for the smart walker robot considering two different maps, where our approach achieved satisfactory results related to safety, execution time and path length.