Dynamic variations in task popularity and resource states tend to reduce cache resource utilization and cache hit rate in edge environments, while increasing system response latency. To address these issues, an adaptive edge caching strategy incorporating dynamic redundant coding (DR-GC) was proposed. A constrained convex optimization objective function was constructed through a task-oriented dynamic redundant coding mechanism, and dual theory was employed to dynamically adjust the coding granularity of tasks. An adaptive caching strategy based on evolutionary game theory was further designed, enabling edge servers to autonomously evolve toward optimal caching strategy combinations in complex environments characterized by limited resources and frequent fluctuations in task requests. Simulation results show that, compared with ARC, PPCS, PaCC, and PFEdge, DR-GC improves the average cache hit rate by approximately 14.5% and reduces the average response latency by approximately 56.3%. The proposed strategy also achieved superior performance in key metrics, including hit time, miss time, cache replacement frequency, and backhaul traffic.