This paper presents a new VLSI-efficient algorithm for robotic exploration in a dynamic environment where the geometry of the objects or their motion trajectories are not known a priori. The input to the proposed algorithm is a list of G nodes obtained using the robot's step size and the dimensions of the environment. P nodes accessible to the robot are identified. The time complexity of the proposed algorithm is O(G). Special features of the algorithm include parallel processing of data from multiple ultrasonic sensors and the use of associative memory to efficiently keep track of the visited terrain nodes. A novel architecture based on selective shutdown of hardware modules for reducing energy consumption is proposed. Detailed experiments with a mobile robot fabricated locally with a Xilinx XC2S200E field programmable gate array and eight ultrasonic sensors on-board validate the efficacy of the proposed approach.