In this paper, we investigate consensus and disturbance attenuation in a chain of mobile agents, which include non-autonomous agents, semi-autonomous agents and autonomous agents. In particular, the nonlinear dynamics of non-autonomous agents is given and cannot be designed, while the dynamics of semi-autonomous and autonomous agents can be partially and fully designed, respectively. To improve the robustness of multi-agent chains against disturbances, we propose a nonlinear control framework for semi-autonomous and autonomous agents such that they mimic the behavior of non-autonomous agents for compatibility while also exploiting long-range connections with distant agents. This framework ensures the existence of a unique consensus equilibrium, which is independent of the network size, connectivity topologies, control gains and information delays. Robustness of multi-agent chains against disturbances is investigated by evaluating the frequency response at the nonlinear level. For infinitely long multi-agent chains with recurrent patterns, we also derive a condition that ensures the disturbance attenuation but only requires the analysis of the linearized model. A case study is conducted for a connected vehicle system where numerical simulations are used to validate the analytical results. disturbance is amplified while propagating along the chain of vehicles [14,15]. Disturbance attenuation in undirected networks of agents with identical linear dynamics was investigated in [16], while a distributed H 1 control for network consensus was presented in [17]. For chains of connected and automated vehicles (CAVs), disturbance attenuation is often called 'string stability' and has been widely studied [18][19][20][21][22][23][24][25].The aforementioned studies on disturbance attenuation in networks assumed that the dynamics of all agents can be designed. However, in practice, there may exist non-autonomous agents (NAAs) that follow certain rules based on their own perception so that their dynamics cannot be designed. On the other hand, the dynamics of semi-autonomous (SAAs) and autonomous (AAs) agents may be partially and fully designed, respectively, while they may also exploit long-range interactions with distant agents. For example, this occurs in connected vehicle systems where human-driven vehicles are mixed with vehicles of higher levels of autonomy that can exploit wireless vehicle-to-vehicle communication [3]. Similar phenomena can be found when attaching controller genes to gene regulatory networks [26] and when controlling neural ensembles using brain-machine interfaces [27]. In nature, the dynamics of NAAs is often nonlinear. For compatibility, it is crucial to ensure that the SAAs and AAs follow similar rules as the NAAs. Thus, their controllers need to be nonlinear as well. Moreover, time delays often arise in the information exchange between agents. Distributed nonlinear control for consensus and disturbance attenuation in time-delayed networks that include NAAs is still an open problem.In this paper, we foc...