Efficient and eco-friendly sustainable waste collection is a growing issue in large scale cities which in turn see the fixed route collection systems to cause issues of fuel efficiency, slow response times, and of course overflowing bins. This work suggests a smart waste collection system founded on Multi Agent Systems (MAS), in which we put forward smart bins and garbage trucks as agents which can think and act for themselves and which also communicate and coordinate in real time. Bin agents which are fitted with IoT sensors report on fill levels and put out collection requests, at the same time truck agents which are supported by GPS and hybrid routing which in turn designed using heuristic A* search combined with a Q-learning reinforcement model for dynamic route planning. Simulation experiments on an urban model comprising of 100 cells and 10 vehicles show that the MAS-based approach leads to 25% less fuel consumption, 88% few overflows, and 40% faster response times compared to the STATIC routing. In addition to the fuel savings, the approach also helps achieve several United Nations Sustainable Development Goals (SDGs): improved urban health and sanitation (SDG 3), promotion of innovative smart infrastructures (SDG 9), improved the provision of integrated and sustainable urban services (SDG 11), efficient resource consumption (SDG 12), and lessened the adverse impacts of climate change through emission reduction (SDG 13). The results affirmed and collaborated the role of AI and IoT enabled MAS systems in improving operational effectiveness while also contributing towards the sustainable development of smart cities.