A Goal-Oriented Approach for Active Object Detection with Exploration-Exploitation Balance

Abstract

Active object detection, which aims to identify objects of interest through controlled camera movements, plays a pivotal role in real-world visual perception for autonomous robotic applications, such as manufacturing tasks (e.g., assembly operations) performed in unknown environments. A dual control for exploration and exploitation (DCEE) algorithm is presented within goal-oriented control systems to achieve efficient active object detection, leveraging active learning by incorporating variance-based uncertainty estimation in the cost function. This novel method employs an exploration-exploitation balanced cost function to actively guide the selection of the next viewpoint. Specifically, active object detection is achieved through the development of a reward function that encodes knowledge about the confidence variation of objects as a function of viewpoint position within a given domain. By identifying the unknown parameters of this function, the system generates an optimal viewpoint planning strategy. DCEE integrates parameter estimation of the reward function and view planning, ensuring a balanced trade-off between the exploitation of learned knowledge and active exploration during the planning process. Moreover, it demonstrates remarkable adaptability across diverse scenarios, effectively handling LEGO brick detection at varying locations. Importantly, the algorithm maintains consistent configuration settings and a fixed number of parameters across various scenarios, underscoring its efficiency and robustness. To validate the proposed approach, extensive numerical studies, high-fidelity virtual simulations, and real-world experiments under various scenarios were conducted. The results confirm the effectiveness of DCEE in active object detection, showcasing superior performance compared to existing methods, including model predictive control (MPC) and entropy approaches.

Figures (14)

• Figure 1: An example of active object detection using a universal robot equipped with a camera is illustrated. The red star marks the terminal position, where the confidence score of the object of interest (i.e., the red cube) exceeds a predefined threshold (e.g., 0.9). The blue area represents an obstacle obstructing the cube from one side. The numerical values at each position indicate the corresponding confidence scores of the object.
• Figure 2: Structure of DCEE-based goal-oriented control systems (GOCS) for active object detection, where $C_k = C(\mathrm{p}_k,\theta)$.
• Figure 3: A virtual environment is set up in Isaac Sim, featuring a red $2\times2$ LEGO brick, with an obstacle positioned on the negative side of the $y$ axis indicated by a green arrow pointing away from the obstacle (i.e., the white area). This setup is designated as Scenario 1 (S1).
• Figure 4: Collected datasets and the generated reward function with parameters listed in Table \ref{['table_6para_carte_sys_lego_s1']}, illustrate the confidence scores of the red LEGO brick (i.e., the red cube) detected using the YOLOv5-s model for S1, as shown in Fig. \ref{['fig_dataset_env_lego_s1_aod']}. The data is acquired within a hemispherical domain, divided into an $11\times21$ grid. The blue area represents the obstacle.
• Figure 5: Collected datasets and the generated reward function with parameters listed in Table \ref{['table_6para_carte_sys_lego_s2']}, illustrate the confidence scores of a red $2 \times 2$ LEGO brick for S2. The data is acquired within a hemispherical domain, divided into an $11\times21$ grid. The blue area represents the obstacle.

Theorems & Definitions (11)

• Definition 1
• Remark 1
• Remark 2
• Remark 3
• Remark 4
• Remark 5
• Remark 6
• Remark 7
• Remark 8
• Remark 9