Disruption-tolerant network (DTN) implementation is subject to many routing constraints like limited knowledge of the network and intermittent connections with no end-to-end path existence. In this paper, the researchers propose trusted-cluster-based routing protocol (TCR) for routing in DTN. TCR uses the experiential learning model that integrates neural network-based bipolar sigmoid activation function to form trusted-cluster DTN. TCR works in two phases: firstly to form a trusted-cluster and secondly to identify cluster heads to direct network traffic through them. After the formation of the trusted-cluster, a cluster head is chosen for a set period, thus instigating stability in the network. These trust values are attached to the node's route cache to make competitive routing decisions by relaying a message to the other trusted intermediate nodes only. With negative trust value, any node is deprived of participation in the network. This way, TCR eliminates malicious or selfish nodes to participate in the DTN network and minimizes the number of messages forwarded in a densely populated DTN. Also, this implementation conserves sufficient buffer memory to reach the destined node. The TCR's performance with other DTN routing schemes, namely, epidemic and trust-based routing, is compared using multiple simulations runs. The proposed work is verified using mobility traces from Community Resource for Archiving Wireless Data At Dartmouth, and the experimental result shows the elimination of selfish nodes participating in the DTN. The simulation result shows an increase of 19% in message delivery by forwarding only to a trusted intermediate node possible. K E Y W O R D S bipolar sigmoid activation function, delay-tolerant network, experiential learning model, intermittently connected network, optimization, trust-based routing