Octopuses control their multiple, soft limbs with the aid of sophisticated peripheral neural circuitry within their arms and suckers. The octopus is largely dependent on this peripheral nervous system and the information acquired by their densely innervated suckers to navigate and forage. Characterizing how these local neural circuits locally generate adaptive behavior within soft limbs provides an approach for the development of technologies involving soft robotics, distributed computing, and neuroprosthetics. Octopuses commonly forage at night and reach their arms into visually occluded spaces while searching for prey. To characterize the octopus’ ability to forage using only the local chemotactile systems within its arms, we investigated the strategies the Pacific red octopus (Octopus rubescens) uses to find food using a single arm within a visually occluded environment. We developed these environments using computer-aided design and they were 3D printed for experimental use. We trained octopuses to use a single arm to explore these 3D printed environments for a food reward. By varying the location of the food item within these environments and tracking the arm using DeepLabCut markerless pose estimation software, we characterized movement patterns used by the arm as it foraged. We used a simplified compartmental model of the arm to evaluate control algorithms that could support these movement patterns. Our results suggest that, absent visual guidance, the octopus relies on a collision-based arm control strategy that emerges from mechanisms of sucker coordination to simplify the control of its soft, highly flexible limbs.