Under the pressure of global warming and depletion of Earth natural resources, the increasing willingness to produce sustainable goods determines us to stop thinking linearly and to shift towards a circular approach by closing material loops. Consistent with these global trends, the industrial symbiosis seeks to generate sustainable advantages by binding traditionally separate industrial processes into a joint production approach, involving physical exchange of by-products or other collateral resources. This thesis introduces new production planning problems raised within industrial symbiosis networks, namely: (i) at the production unit level: a single-item lot-sizing problem with inventory capacities, which integrates the management of by-products, and (ii) network-wide: a two-level lot-sizing problem, which aims to align lot-sizing decisions of two actors: one supplier of by-products and its receiver. The complexity of the studied problems is analyzed and efficient algorithms based on dynamic programming and Lagrangian decomposition are proposed. To support the framing of symbiotic partnerships within an industrial symbiosis network, we investigate several collaboration policies for different levels of information sharing: (i) centralized and decentralized collaboration policies based on mixed-integer linear programming, for full and none information sharing, (ii) a game-theoretic collaboration policy for asymmetric information sharing, and (iii) a contractual-based collaboration policy obtained via a negotiation-based scheme managed by a blinded mediator, for symmetric feedback sharing. Through extensive numerical experiments, the proposed collaboration policies are discussed according to three dimensions: the satisfaction of actors, environmental implications and economic benefits.