COBOT:A Software Framework For Collaborative Self-Adaptive Systems

Self-adaptive systems are able to adjust their behavior in response to perception of their environments, and smartly complete their tasks in an autonomic way. Compared with single self-adaptive system, applications based on multiple self-adaptive systems usually possess more excellent features, such as better parallelism, fault tolerance and so on. But at the same time, design and development of these applications are subject to new challenges, we found that probably the most difficult challenges lies in the following two aspects:● Universality of system. Besides their own environment conditions, systems in distributed environment also need to consider other systems’ strategies for de-cision making. Strategies may vary wildly for different application entities, d-ifferent number of entities and different missions. Thus we need to modify the logic of context awareness, context analysis, self-adaptive decision and system implementation. However, conducting such modifications is difficult and may introduce new bugs due to the complexity of multiple systems. Thus program-mers may prefer re-developing instead of modifying existing code, which would consume a lot of human resources.● Reliability of system. One remarkable feature of self-adaptive system is auto-maticity. After initialized, systems can automatically conduct analysis and make decisions without human intervention. However, self-adaptive systems might be interfered by failure of sensors, actuators and other system components, which would finally lead to execution anomaly or even system failure.Development of multiple self-adaptive applications are difficult and error-producing due to the lack of universality and reliability. We in this article address this research issue from the following aspects:● To solve the universality problem, we propose a software framework COBOT for multiple self-adaptive systems based on CO-SCAS rule model. Self-adaptation logic is the key of self-adaptive systems. Usually, such adaptation logic is speci-fied based on some rule model. Thus a good rule model can essentially solve the universality problem. We in this article proposed a rule model called CO-SCAS (Coordination State-Condition-Action-State) based on A-FSM (Adaptive Finite-State Machine) and DEC-POMDF (DECentralized Partially Observable Markov Decision Process Model). On the basis of CO-SCAS, we propose our COBOT framework. COBOT follows the principle of Separation of Concern and divides components into multiple layers for decoupling.● To solve the problem of reliability, we design a CCR (Context Check and Re-covery) module for the framework. There have been existing research works on the reliability and QoS of self-adaptive systems. Most of these techniques conduct fallback or compensation operations after anomaly detected. However, such operations can be prohibited in some situations. We in this article designed a fault-tolerant module called CCR, which improves the reliability by preventing anomaly in context analysis layer affecting the self-adaptive layer.We conduct experiments to evaluate the effectiveness of COBOT. A Robot-Car application, where robot cars are designed to collaboratively explore unknown environ-ment, is used as the subject. Both simulated and real-world experiments are included. Experimental results confirmed the universality and reliability of COBOT framework.