In this study, we use the type-2 fuzzy neural controller (T2FNC) based on the improved mantis search algorithm (IMSA) for navigation and obstacle avoidance applications of unmanned ground vehicles (UGVs) with differential wheels in unknown environments. In unknown environments, a light detection and ranging sensor is used to capture distance information between UGVs and the surrounding environment. The T2FNC has a five-layer architecture. The first to fifth layers are the input, fuzzified, rule, order reduction process, and output layers, respectively. In the T2FNC, we use the IMSA to adjust the parameters in the network. In addition, the simulated annealing reciprocal local search algorithm is proposed to prevent the traditional MSA from falling into the local optimal solution. Experimental results indicate that the fitness value of the proposed IMSA is 0.983452. Compared with the traditional MSA algorithm, the movement distance and movement time of the proposed T2FNC with IMSA are shortened by 4.5 and 4.14%, respectively. In addition, the experimental results show that the proposed method can have excellent obstacle avoidance and navigation capabilities in unknown environments.