The last decade has seen a shift in emphasis from centralised to decentralised systems to meet the demanding coordination requirements of today's complex computer systems. In such systems, the aim is to achieve effective decentralised control through autonomous software agents that perform local decision-making based on incomplete and imperfect information. Specifically, when the various agents interact, the system behaves as a computational ecology with no single agent coordinating their actions. In this thesis, we focus on one specific type of computational ecology, the Continuous Double Auction (CDA), and investigate market-oriented approaches to decentralised control. In particular, the CDA is a fixed-duration auction mechanism where multiple buyers and sellers compete to buy and sell goods, respectively, in the market, and where transactions can occur at any time whenever an offer to buy and an offer to sell match. Now, in such a market mechanism, the decentralised control is achieved through the decentralised allocation of resources, which, in turn, is an emergent behaviour of buyers and sellers trading in the market. The CDA was chosen, among the plenitude of auction formats available, because it allows efficient resource allocation without the need of a centralised auctioneer. Against this background, we look at both the structure and the behaviour of the CDA in our attempt to build an efficient and robust mechanism for decentralised control. We seek to do this for both stable environments, in which the market demand and supply do not change and dynamic ones in which there are sporadic changes (known as market shocks). While the structure of the CDA defines the agents' interactions in the market, the behaviour of the CDA is determined by what emerges when the buyers and sellers compete to maximise their individual profits. In more detail, on the structural aspect, we first look at how the market protocol of the CDA can be modified to meet desirable properties for the system (such as high market efficiency, fairness of profit distribution among agents and market stability). Second, we use this modified protocol to efficiently solve a complex decentralised task allocation problem with limited-capacity suppliers that have start-up production costs and consumers with inelastic demand. Furthermore, we demonstrate that the structure of this CDA variant is very efficient (an average of 80% and upto 90%) by evaluating the mechanism with very simple agent behaviours. In so doing, we emphasise the effect of the structure, rather than the behaviour, on efficiency. In the behavioural aspect, we first developed a multi-layered framework for designing strategies that autonomous agents can use for trading in various types of market mechanisms. We then use this framework to design a novel Adaptive-Aggressiveness (AA) strategy for the CDA. Specifically, our bidding strategy has both a short and a long-term learning mechanism to adapt its behaviour to changing market conditions and it is designed to be robust in both static and dynamic environments. Furthermore, we also developed a novel framework that uses a two-population evolutionary game theoretic approach to analyse the strategic interactions of buyers and sellers in the CDA. Finally, we develop effective methodologies for evaluating strategies for the CDA in both homogeneous and heterogeneous populations, within static and dynamic environments. We then evaluate the AA bidding strategy against the state of the art using these methodologies. By so doing, we show that, within homogeneous populations, the AA strategy outperformed the benchmarks, in terms of market efficiency, by up to 3.6% in the static case and 2.8% in the dynamic case. Within heterogeneous populations, based on our evolutionary game theoretic framework, we identify that there is a probability above 85% that the AA strategy will eventually be adopted by buyers and sellers in the market (for being more efficient) and, therefore, AA is also better than the benchmarks in heterogeneous populations as well.