This paper studies congestion avoidance in a simple two-dimensional traffic system, using computer simulation. The mobile objects avoid congestion among themselves using simple congestion-avoiding traffic rules. The objects adaptively avoid congested regions and move towards less congested regions. The objects avoid congestion in decentralized manner based only on congestion levels in their local regions. It is found that the adaptive decentralized congestion-avoiding traffic rules prevent the traffic from undergoing congestion phase transition at low critical density. The congestion avoidance significantly increases the traffic capacity. The congestion-avoiding traffic rules increase the traffic capacity by keeping the emerging congestion and traffic hot spots small, localized, and temporary. Due to congestion avoidance, the travel time of objects is high and the amount of flow is low. The congestion-avoiding traffic eventually undergoes phase transition from free flow to jammed state, but at high critical density.